forked from mtusi2026/mtusi2026lab4
перенос: реализация lab3 без изменений (API, миграции 0001-0003, HAProxy, Valkey, тесты)
Исходники, Dockerfile, compose.yaml, скрипты и собственные тесты скопированы из mtusi2026lab3 (ветка solution/bvt2302-bodanyuk, b6847e2) как есть; изменения lab4 — отдельными коммитами. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5.5
parent
ec540a1a72
commit
d916d09f7b
+29
-8
@@ -5,16 +5,37 @@ APP_PORT=8080
|
||||
APP_ENV=lab
|
||||
SESSION_TTL_SECONDS=3600
|
||||
# Choose credentials locally. File paths contain no secret values.
|
||||
# The files are created by `python scripts/init_secrets.py` and mounted by compose as secrets.
|
||||
DATABASE_URL_FILE=/run/secrets/database_url
|
||||
# Optional on 3, required use on 4 in lab2; Redis OR Valkey:
|
||||
SESSION_STORE=postgres
|
||||
# valkey = sessions live only in Valkey (AOF, appendfsync always); postgres = table `sessions`.
|
||||
SESSION_STORE=valkey
|
||||
CACHE_URL_FILE=/run/secrets/cache_url
|
||||
# Number of interchangeable API replicas behind HAProxy (compose deploy.replicas).
|
||||
API_REPLICAS=2
|
||||
# Idempotency-Key results are kept at least this long (contract: >= 24 h).
|
||||
IDEMPOTENCY_TTL_SECONDS=86400
|
||||
S3_ENDPOINT=http://s3:9000
|
||||
S3_REGION=us-east-1
|
||||
S3_BUCKET=messenger-attachments
|
||||
S3_ACCESS_KEY_ID_FILE=/run/secrets/s3_access_key_id
|
||||
S3_SECRET_ACCESS_KEY_FILE=/run/secrets/s3_secret_access_key
|
||||
MAX_UPLOAD_BYTES=10485760
|
||||
MAX_IMAGE_PIXELS=40000000
|
||||
# Write path of POST message: function = one SQL call (migration 0003), multi = lab3 baseline path.
|
||||
MESSAGE_WRITE_PATH=function
|
||||
# Additional settings of this implementation (defaults shown; all optional):
|
||||
LOG_LEVEL=info
|
||||
DB_POOL_MIN=1
|
||||
DB_POOL_MAX=10
|
||||
DB_TIMEOUT_SECONDS=3
|
||||
CACHE_TIMEOUT_SECONDS=0.5
|
||||
PASSWORD_HASH_CONCURRENCY=2
|
||||
DOCS_ENABLED=true
|
||||
# Graceful stop of a replica: readiness 503 + drain, then up to SHUTDOWN for in-flight requests.
|
||||
DRAIN_SECONDS=3
|
||||
SHUTDOWN_TIMEOUT_SECONDS=8
|
||||
# CPU/RAM limits (docker compose deploy.resources.limits); API_* are per replica.
|
||||
API_CPUS=1.0
|
||||
API_MEMORY=256m
|
||||
PG_CPUS=2.0
|
||||
PG_MEMORY=1g
|
||||
VALKEY_CPUS=0.5
|
||||
LB_CPUS=1.0
|
||||
PG_MAX_CONNECTIONS=100
|
||||
# Host ports of the load balancer (published on 127.0.0.1 only; API replicas publish nothing)
|
||||
APP_PUBLISH_PORT=8080
|
||||
LB_STATS_PORT=8404
|
||||
|
||||
@@ -0,0 +1,5 @@
|
||||
# LF everywhere: identical files on Windows/Linux and inside `git archive | docker build -`.
|
||||
* text=auto eol=lf
|
||||
*.png binary
|
||||
# cmd.exe expects CRLF in batch files.
|
||||
*.cmd text eol=crlf
|
||||
@@ -0,0 +1,36 @@
|
||||
name: Application (lab3)
|
||||
on: [push, pull_request]
|
||||
permissions:
|
||||
contents: read
|
||||
jobs:
|
||||
app:
|
||||
runs-on: ubuntu-latest
|
||||
timeout-minutes: 40
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: actions/setup-python@v5
|
||||
with:
|
||||
python-version: '3.13'
|
||||
- name: Synthetic secrets and env (CI only, never committed)
|
||||
run: python scripts/init_secrets.py && cp .env.example .env
|
||||
- name: Build, migrate once, start 2 API replicas behind HAProxy, wait for healthy
|
||||
run: docker compose --env-file .env up -d --build --wait
|
||||
- name: Contract and public smoke
|
||||
run: make check test PYTHON=python
|
||||
- name: Persistence across restart of LB, API replicas and stores
|
||||
run: make test-persistence ACTION_SCRIPT=scripts/restart.sh PYTHON=python
|
||||
- name: Failover of the replica that served the test, then restore it
|
||||
run: |
|
||||
make test-failover ACTION_SCRIPT=scripts/stop-one.sh PYTHON=python
|
||||
sh scripts/restore-api.sh && python scripts/lbctl.py wait-up 2
|
||||
- name: Own tests (unit + stand; every failure is restored in finally)
|
||||
run: |
|
||||
python -m pip install --require-hashes --no-deps -r requirements.txt
|
||||
python -m pip install -r requirements-test.txt
|
||||
make student-test PYTHON=python
|
||||
- name: Logs
|
||||
if: always()
|
||||
run: docker compose logs --tail 100 api lb
|
||||
- name: Remove CI stand and its synthetic data
|
||||
if: always()
|
||||
run: docker compose down -v
|
||||
+59
@@ -0,0 +1,59 @@
|
||||
# syntax=docker/dockerfile:1
|
||||
# Messenger API, lab 3 (replicas behind HAProxy, PostgreSQL + Valkey). Built by compose.yaml (services api, migrate).
|
||||
#
|
||||
# Python is interpreted, so there is no separate compile stage; the image
|
||||
# contains only the interpreter, locked dependencies and the source code.
|
||||
# Base image is pinned by version AND digest (python 3.13.16, Debian 12 bookworm).
|
||||
FROM python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
|
||||
|
||||
ENV PYTHONDONTWRITEBYTECODE=1 \
|
||||
PYTHONUNBUFFERED=1 \
|
||||
PIP_NO_CACHE_DIR=1 \
|
||||
PIP_DISABLE_PIP_VERSION_CHECK=1
|
||||
|
||||
WORKDIR /app
|
||||
|
||||
# 1. Dependencies first (layer cache survives source edits). Every package,
|
||||
# including transitive ones, is pinned with sha256 hashes in requirements.txt;
|
||||
# --no-deps guarantees nothing outside the lock file is installed.
|
||||
COPY requirements.txt ./
|
||||
RUN pip install --root-user-action=ignore --require-hashes --no-deps -r requirements.txt
|
||||
|
||||
# 2. Unprivileged runtime user. Code stays owned by root, so the process
|
||||
# cannot modify its own code.
|
||||
RUN groupadd --system --gid 10001 app \
|
||||
&& useradd --system --uid 10001 --gid app --no-create-home --shell /usr/sbin/nologin app
|
||||
|
||||
COPY src/ ./src/
|
||||
# Versioned SQL migrations, applied by `python -m messenger.migrate` (compose service migrate).
|
||||
COPY migrations/ ./migrations/
|
||||
# Course contract, served read-only at /openapi.json for Swagger UI (/docs).
|
||||
COPY contracts/openapi.json ./contracts/openapi.json
|
||||
|
||||
# Non-secret defaults; compose.yaml passes the rest. Passwords come only as files
|
||||
# (DATABASE_URL_FILE / CACHE_URL_FILE -> Docker secrets), never as ENV or COPY.
|
||||
ENV PYTHONPATH=/app/src \
|
||||
APP_HOST=0.0.0.0 \
|
||||
APP_PORT=8080 \
|
||||
APP_ENV=lab \
|
||||
LOG_LEVEL=info \
|
||||
SHUTDOWN_TIMEOUT_SECONDS=8 \
|
||||
DOCS_ENABLED=true \
|
||||
SESSION_TTL_SECONDS=3600 \
|
||||
DATABASE_URL_FILE=/run/secrets/database_url
|
||||
|
||||
# Build revision (scripts/rolling-update.sh passes git commit + time): makes every rolled-out
|
||||
# image distinguishable in `docker inspect` and in the "starting" log line of each replica.
|
||||
ARG BUILD_REVISION=dev
|
||||
LABEL org.opencontainers.image.revision=$BUILD_REVISION
|
||||
ENV APP_REVISION=$BUILD_REVISION
|
||||
|
||||
USER app:app
|
||||
EXPOSE 8080
|
||||
STOPSIGNAL SIGTERM
|
||||
|
||||
HEALTHCHECK --interval=10s --timeout=3s --start-period=5s --retries=3 \
|
||||
CMD ["python", "-m", "messenger.healthcheck"]
|
||||
|
||||
# Exec form: the Python server is PID 1 and receives SIGTERM directly (no shell in between).
|
||||
CMD ["python", "-m", "messenger"]
|
||||
@@ -28,3 +28,55 @@ test-persistence:
|
||||
test-failover:
|
||||
@test -n "$(ACTION_SCRIPT)" || (echo "Set ACTION_SCRIPT to your implemented executable stop-one hook"; exit 2)
|
||||
$(PYTHON) tests/resilience.py --mode failover --action-script "$(ACTION_SCRIPT)" --base-url "$(BASE_URL)" $(TLS_ARGS)
|
||||
# ---------------------------------------------------------------------------
|
||||
# Student solution (lab 3, Python). On Windows: PowerShell/Git Bash with GNU make, public
|
||||
# tests need PYTHONUTF8=1, and the hooks are started through scripts/*.cmd (Git Bash).
|
||||
.PHONY: secrets up down logs ps migrate student-test unit-test explain lb-status restore rolling-update failover-series
|
||||
|
||||
secrets:
|
||||
$(PYTHON) scripts/init_secrets.py
|
||||
|
||||
# Build, apply migrations ONCE (service migrate), start API replicas and the LB, wait for healthy.
|
||||
up: secrets
|
||||
docker compose up -d --build --wait
|
||||
|
||||
# Safe stop: containers stop, named volumes (data) stay.
|
||||
down:
|
||||
docker compose down
|
||||
|
||||
logs:
|
||||
docker compose logs --tail 50 api lb
|
||||
|
||||
ps:
|
||||
docker compose ps -a
|
||||
|
||||
migrate:
|
||||
docker compose run --rm migrate
|
||||
|
||||
# HAProxy view of the replicas (UP/DOWN) mapped to containers and X-Instance-Id.
|
||||
lb-status:
|
||||
$(PYTHON) scripts/lbctl.py status
|
||||
$(PYTHON) scripts/lbctl.py replicas
|
||||
|
||||
# Start replicas stopped by scripts/stop-one.sh again and wait until the LB has them all UP.
|
||||
restore:
|
||||
sh scripts/restore-api.sh
|
||||
$(PYTHON) scripts/lbctl.py wait-up $${API_REPLICAS:-2}
|
||||
|
||||
# Replace every API replica one by one with a freshly built image (no downtime).
|
||||
rolling-update:
|
||||
sh scripts/rolling-update.sh
|
||||
|
||||
# Level 4: stop each replica in turn under load, measure RPS/errors/recovery (evidence/failover/).
|
||||
failover-series:
|
||||
$(PYTHON) bench/failover_series.py --output evidence/failover
|
||||
|
||||
# Own tests: unit + black-box against the running stand (needs requirements*.txt in a venv).
|
||||
student-test:
|
||||
$(PYTHON) -m pytest student-tests -v
|
||||
|
||||
unit-test:
|
||||
$(PYTHON) -m pytest student-tests -m "not stack" -q
|
||||
|
||||
explain:
|
||||
$(PYTHON) scripts/explain_history.py --output evidence/explain-history.txt
|
||||
|
||||
@@ -0,0 +1,141 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Level 4: stop EACH API replica in turn under load (restoring the previous one first) and
|
||||
measure RPS, errors and recovery. Modes: graceful `stop` (SIGTERM + drain), `kill` (SIGKILL)
|
||||
and `pause` (frozen process: the kernel still accepts TCP, nothing answers - the worst case
|
||||
for detection, only the active health check can find it).
|
||||
|
||||
python bench/failover_series.py --output evidence/failover # ~5 min for 2 replicas x 3 modes
|
||||
|
||||
Timeline of one round (all times from the moment the stop command is issued):
|
||||
warm-up 6 s -> stop replica -> 12 s with it down -> start it again -> wait until HAProxy
|
||||
has every replica UP (rejoin time) -> 6 s steady state.
|
||||
Measured per round: requests/ok/failed and ok-RPS before / during / after, the last failed
|
||||
request (+s), the first moment after which the stopped replica never answered again
|
||||
(= removal from rotation), time until the restarted replica is UP in HAProxy and serves
|
||||
traffic again, plus the history check (no confirmed message lost, no duplicate).
|
||||
Output: <output>/series.json (summary), <output>/timeline-<mode>-<replica>.csv (per second),
|
||||
<output>/series.md (table for REPORT.md). No tokens are written.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import csv
|
||||
import datetime as dt
|
||||
import json
|
||||
import platform
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
sys.path.insert(0, str(ROOT / "scripts"))
|
||||
sys.path.insert(0, str(ROOT / "bench"))
|
||||
import lbctl # noqa: E402
|
||||
from loadgen import Load # noqa: E402
|
||||
|
||||
|
||||
def docker(*args: str) -> str:
|
||||
return subprocess.run(["docker", *args], capture_output=True, text=True, check=True, timeout=120).stdout
|
||||
|
||||
|
||||
def one_round(replica: dict, mode: str, replicas: int, workers: int, out: Path) -> dict:
|
||||
load = Load(workers).start()
|
||||
time.sleep(6)
|
||||
t_stop = load.now()
|
||||
if mode == "pause":
|
||||
docker("pause", replica["id"])
|
||||
else:
|
||||
docker("stop", "--time", "15" if mode == "stop" else "0", replica["id"])
|
||||
t_stopped = load.now()
|
||||
time.sleep(12)
|
||||
t_start = load.now()
|
||||
docker("unpause" if mode == "pause" else "start", replica["id"])
|
||||
lbctl.wait_up(replicas, timeout=90)
|
||||
t_up = load.now()
|
||||
# The restarted container keeps its hostname: wait until it serves requests again.
|
||||
deadline = time.monotonic() + 30
|
||||
while time.monotonic() < deadline:
|
||||
if any(e[3] == replica["hostname"] and e[0] > t_up for e in load.events):
|
||||
break
|
||||
time.sleep(0.1)
|
||||
t_serves = load.now()
|
||||
time.sleep(6)
|
||||
load.stop()
|
||||
t_end = load.now()
|
||||
|
||||
failed_after_stop = [e for e in load.events if e[0] >= t_stop and e[2] not in (200, 201)]
|
||||
served_by_target = [e[0] for e in load.events if e[3] == replica["hostname"] and t_stop <= e[0] < t_start]
|
||||
out_of_rotation = (max(served_by_target) - t_stop) if served_by_target else 0.0
|
||||
rows = load.per_second()
|
||||
with (out / f"timeline-{mode}-{replica['name']}.csv").open("w", newline="", encoding="utf-8") as f:
|
||||
writer = csv.DictWriter(f, fieldnames=["second", "ok", "failed", "instances"])
|
||||
writer.writeheader()
|
||||
for row in rows:
|
||||
row = dict(row)
|
||||
row["instances"] = row["instances"].replace(replica["hostname"], "TARGET")
|
||||
writer.writerow(row)
|
||||
state = json.loads(docker("inspect", replica["id"]))[0]["State"]
|
||||
return {
|
||||
"mode": mode, "replica": replica["name"], "instance_id": replica["hostname"],
|
||||
"stop_command_seconds": round(t_stopped - t_stop, 2),
|
||||
"before": load.window(0.5, t_stop), "during": load.window(t_stop, t_start), "after": load.window(t_serves, t_end),
|
||||
"failed_after_stop": len(failed_after_stop),
|
||||
"failed_statuses": sorted({e[2] for e in failed_after_stop}),
|
||||
"last_failure_after_stop_s": round(max((e[0] for e in failed_after_stop), default=t_stop) - t_stop, 3),
|
||||
"target_last_served_after_stop_s": round(out_of_rotation, 3),
|
||||
"rejoin_up_in_lb_s": round(t_up - t_start, 2), "rejoin_serving_s": round(t_serves - t_start, 2),
|
||||
"client_retries_same_key": load.retries,
|
||||
"history": load.verify_history(),
|
||||
"restarted_state": f"{state['Status']} health={(state.get('Health') or {}).get('Status')}",
|
||||
}
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
|
||||
parser.add_argument("--output", default="evidence/failover")
|
||||
parser.add_argument("--modes", default="stop,kill,pause")
|
||||
parser.add_argument("--workers", type=int, default=4)
|
||||
args = parser.parse_args()
|
||||
out = ROOT / args.output
|
||||
out.mkdir(parents=True, exist_ok=True)
|
||||
replicas = sorted((r for r in lbctl.replicas() if r["state"] == "running"), key=lambda r: r["name"])
|
||||
assert len(replicas) >= 2, "need at least two running replicas"
|
||||
lbctl.wait_up(len(replicas))
|
||||
results = []
|
||||
for mode in args.modes.split(","):
|
||||
for replica in replicas: # each replica in turn; the previous one is already restored
|
||||
print(f"== {mode} {replica['name']} ...", flush=True)
|
||||
result = one_round(replica, mode, len(replicas), args.workers, out)
|
||||
results.append(result)
|
||||
print(json.dumps({k: result[k] for k in ("failed_after_stop", "last_failure_after_stop_s",
|
||||
"target_last_served_after_stop_s", "rejoin_serving_s")}), flush=True)
|
||||
meta = {"date": dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"),
|
||||
"commit": subprocess.run(["git", "rev-parse", "--short", "HEAD"], cwd=ROOT, capture_output=True,
|
||||
text=True).stdout.strip(),
|
||||
"host": f"{platform.system()} {platform.release()}, docker {docker('version', '--format', '{{.Server.Version}}').strip()}",
|
||||
"workload": f"{args.workers} closed-loop workers from the host via 127.0.0.1:8080: POST message "
|
||||
"(Idempotency-Key, uncertain -> same key) + GET last page; client timeout 3 s",
|
||||
"lb": "HAProxy: check GET /health/ready inter 1s fastinter 500ms fall 2 rise 2, timeout check 1s; "
|
||||
"retry-on conn-failure + redispatch",
|
||||
"api": "DRAIN_SECONDS=3, SHUTDOWN_TIMEOUT_SECONDS=8, stop_grace_period 15s"}
|
||||
(out / "series.json").write_text(json.dumps({"meta": meta, "rounds": results}, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8")
|
||||
lines = ["| Режим | Реплика | Запросов во время отказа | Ошибок после stop | ok-RPS до / во время / после | "
|
||||
"Последняя ошибка, с | Цель ушла из ротации, с | Возврат в LB / обслуживает, с | Потери / дубли |",
|
||||
"| --- | --- | --- | --- | --- | --- | --- | --- | --- |"]
|
||||
for r in results:
|
||||
h = r["history"]
|
||||
lines.append(f"| {r['mode']} | {r['replica']} | {r['during']['requests']} | {r['failed_after_stop']} "
|
||||
f"{r['failed_statuses'] or ''} | {r['before']['ok_rps']} / {r['during']['ok_rps']} / {r['after']['ok_rps']} | "
|
||||
f"{r['last_failure_after_stop_s']} | {r['target_last_served_after_stop_s']} | "
|
||||
f"{r['rejoin_up_in_lb_s']} / {r['rejoin_serving_s']} | "
|
||||
f"{h['lost_confirmed']} / {h['duplicate_ids'] + h['duplicate_texts']} |")
|
||||
(out / "series.md").write_text("\n".join([f"Дата {meta['date']}, commit {meta['commit']}. {meta['workload']}.", "",
|
||||
*lines, ""]), encoding="utf-8")
|
||||
print("\n".join(lines))
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1,212 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Background HTTP traffic through the LB for failover / rolling-update experiments (level 4).
|
||||
|
||||
Closed loop, N worker threads, each: POST a message (Idempotency-Key) + GET the last page.
|
||||
An uncertain POST (timeout, connection error, 5xx) is retried with the SAME key and body,
|
||||
exactly as a correct client must do; the history is checked afterwards for losses and
|
||||
duplicates. Every request is recorded as (t, kind, status, instance, latency).
|
||||
This is a correctness/availability probe, not a capacity benchmark (see bench/track1 for that).
|
||||
|
||||
python bench/loadgen.py --duration 30 --workers 4 # prints a summary JSON
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import base64
|
||||
import json
|
||||
import math
|
||||
import os
|
||||
import threading
|
||||
import time
|
||||
import uuid
|
||||
from collections import Counter, defaultdict
|
||||
|
||||
import urllib.error
|
||||
import urllib.request
|
||||
|
||||
BASE_URL = os.environ.get("BASE_URL", "http://127.0.0.1:8080").rstrip("/")
|
||||
|
||||
|
||||
class Http:
|
||||
"""Minimal stdlib client with keep-alive per thread (http.client under urllib is not)."""
|
||||
|
||||
def __init__(self, base_url: str, timeout: float):
|
||||
import http.client
|
||||
from urllib.parse import urlsplit
|
||||
parts = urlsplit(base_url)
|
||||
self.host, self.port = parts.hostname, parts.port or 80
|
||||
self.timeout = timeout
|
||||
self.factory = lambda: http.client.HTTPConnection(self.host, self.port, timeout=self.timeout)
|
||||
self.conn = self.factory()
|
||||
|
||||
def request(self, method: str, path: str, body: bytes | None = None, headers: dict | None = None):
|
||||
try:
|
||||
self.conn.request(method, path, body=body, headers=headers or {})
|
||||
response = self.conn.getresponse()
|
||||
data = response.read()
|
||||
if response.getheader("connection", "").lower() == "close":
|
||||
self.conn.close()
|
||||
self.conn = self.factory()
|
||||
return response.status, dict(response.getheaders()), data
|
||||
except Exception:
|
||||
self.conn.close()
|
||||
self.conn = self.factory()
|
||||
raise
|
||||
|
||||
|
||||
def call(method: str, path: str, body=None, headers=None, timeout=10.0):
|
||||
data = json.dumps(body).encode() if body is not None else None
|
||||
h = {"Content-Type": "application/json", **(headers or {})}
|
||||
req = urllib.request.Request(BASE_URL + path, data=data, headers=h, method=method)
|
||||
try:
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.status, json.loads(r.read() or b"null")
|
||||
except urllib.error.HTTPError as e:
|
||||
return e.code, json.loads(e.read() or b"null")
|
||||
|
||||
|
||||
def make_user(label: str) -> dict:
|
||||
username = f"{label}_{uuid.uuid4().hex[:12]}"
|
||||
password = "Load-password-" + uuid.uuid4().hex
|
||||
status, user = call("POST", "/api/v1/users", {"username": username, "display_name": label, "password": password})
|
||||
assert status == 201, (status, user)
|
||||
cred = base64.b64encode(f"{username}:{password}".encode()).decode()
|
||||
status, session = call("POST", "/api/v1/auth/sessions", headers={"Authorization": "Basic " + cred})
|
||||
assert status == 201, (status, session)
|
||||
return {"id": user["id"], "token": session["token"]}
|
||||
|
||||
|
||||
class Load:
|
||||
def __init__(self, workers: int = 4, timeout: float = 3.0, pause: float = 0.0):
|
||||
self.workers = workers
|
||||
self.timeout = timeout
|
||||
self.pause = pause
|
||||
self.alice, self.bob = make_user("load"), make_user("load")
|
||||
status, chat = call("POST", "/api/v1/chats", {"title": "load", "member_ids": [self.alice["id"], self.bob["id"]]},
|
||||
headers={"Authorization": "Bearer " + self.alice["token"]})
|
||||
assert status == 201, chat
|
||||
self.chat = chat
|
||||
self.stop_event = threading.Event()
|
||||
self.lock = threading.Lock()
|
||||
self.events: list[tuple[float, str, int, str, float]] = []
|
||||
self.confirmed: dict[str, dict] = {}
|
||||
self.retries = 0
|
||||
self.t0 = time.monotonic()
|
||||
self.threads = [threading.Thread(target=self._worker, args=(i,), daemon=True) for i in range(workers)]
|
||||
|
||||
def now(self) -> float:
|
||||
return time.monotonic() - self.t0
|
||||
|
||||
def _record(self, kind: str, status: int, instance: str, latency: float) -> None:
|
||||
with self.lock:
|
||||
self.events.append((round(self.now(), 4), kind, status, instance, round(latency * 1000, 2)))
|
||||
|
||||
def _worker(self, index: int) -> None:
|
||||
http = Http(BASE_URL, self.timeout)
|
||||
auth = {"Authorization": "Bearer " + self.alice["token"]}
|
||||
path = f"/api/v1/chats/{self.chat['id']}/messages"
|
||||
n, key = 0, uuid.uuid4().hex
|
||||
text = f"w{index}-0-{key[:6]}"
|
||||
while not self.stop_event.is_set():
|
||||
started = time.monotonic()
|
||||
try:
|
||||
status, headers, data = http.request("POST", path, json.dumps({"text": text}).encode(),
|
||||
{**auth, "Idempotency-Key": key, "Content-Type": "application/json"})
|
||||
instance = headers.get("x-instance-id") or headers.get("X-Instance-Id") or "-"
|
||||
except Exception:
|
||||
status, instance, data = 0, "-", b""
|
||||
self._record("POST", status, instance, time.monotonic() - started)
|
||||
if status == 201:
|
||||
msg = json.loads(data)
|
||||
with self.lock:
|
||||
self.confirmed[msg["id"]] = msg
|
||||
n += 1
|
||||
key = uuid.uuid4().hex
|
||||
text = f"w{index}-{n}-{key[:6]}"
|
||||
elif status in (0, 502, 503, 504, 500):
|
||||
with self.lock:
|
||||
self.retries += 1 # same key + same text on the next iteration
|
||||
started = time.monotonic()
|
||||
try:
|
||||
status, headers, _ = http.request("GET", path + "?limit=5", None, auth)
|
||||
instance = headers.get("x-instance-id") or headers.get("X-Instance-Id") or "-"
|
||||
except Exception:
|
||||
status, instance = 0, "-"
|
||||
self._record("GET", status, instance, time.monotonic() - started)
|
||||
if self.pause:
|
||||
time.sleep(self.pause)
|
||||
|
||||
def start(self) -> "Load":
|
||||
self.t0 = time.monotonic()
|
||||
for t in self.threads:
|
||||
t.start()
|
||||
return self
|
||||
|
||||
def stop(self) -> None:
|
||||
self.stop_event.set()
|
||||
for t in self.threads:
|
||||
t.join(timeout=15)
|
||||
|
||||
def verify_history(self) -> dict:
|
||||
"""Walk the whole history with the second user: no confirmed message lost, no duplicate."""
|
||||
items, cursor = [], None
|
||||
while True:
|
||||
q = "?limit=100" + (f"&cursor={cursor}" if cursor else "")
|
||||
status, page = call("GET", f"/api/v1/chats/{self.chat['id']}/messages{q}",
|
||||
headers={"Authorization": "Bearer " + self.bob["token"]}, timeout=30)
|
||||
assert status == 200, page
|
||||
items += page["items"]
|
||||
cursor = page["next_cursor"]
|
||||
if not cursor:
|
||||
break
|
||||
ids = [m["id"] for m in items]
|
||||
texts = [m["text"] for m in items]
|
||||
lost = [i for i in self.confirmed if i not in set(ids)]
|
||||
changed = [m for m in items if m["id"] in self.confirmed and self.confirmed[m["id"]] != m]
|
||||
return {"history_messages": len(items), "confirmed_messages": len(self.confirmed),
|
||||
"lost_confirmed": len(lost), "altered_confirmed": len(changed),
|
||||
"duplicate_ids": len(ids) - len(set(ids)), "duplicate_texts": len(texts) - len(set(texts)),
|
||||
"unconfirmed_but_stored": len(set(ids) - set(self.confirmed))}
|
||||
|
||||
# ------------------------------------------------------------------ summaries
|
||||
def window(self, start: float, end: float) -> dict:
|
||||
ev = [e for e in self.events if start <= e[0] < end]
|
||||
ok = [e for e in ev if e[2] in (200, 201)]
|
||||
lat = sorted(e[4] for e in ok)
|
||||
|
||||
def pct(p: float):
|
||||
return lat[max(0, math.ceil(p * len(lat)) - 1)] if lat else None
|
||||
span = max(end - start, 1e-9)
|
||||
return {"seconds": round(span, 2), "requests": len(ev), "ok": len(ok), "failed": len(ev) - len(ok),
|
||||
"ok_rps": round(len(ok) / span, 1), "error_ratio": round((len(ev) - len(ok)) / len(ev), 5) if ev else None,
|
||||
"statuses": dict(Counter(e[2] for e in ev)), "instances": dict(Counter(e[3] for e in ok)),
|
||||
"p50_ms": pct(.5), "p95_ms": pct(.95), "p99_ms": pct(.99)}
|
||||
|
||||
def per_second(self) -> list[dict]:
|
||||
buckets: dict[int, list] = defaultdict(list)
|
||||
for e in self.events:
|
||||
buckets[int(e[0])].append(e)
|
||||
rows = []
|
||||
for second in sorted(buckets):
|
||||
ev = buckets[second]
|
||||
rows.append({"second": second, "ok": sum(1 for e in ev if e[2] in (200, 201)),
|
||||
"failed": sum(1 for e in ev if e[2] not in (200, 201)),
|
||||
"instances": "|".join(f"{k}:{v}" for k, v in sorted(Counter(e[3] for e in ev).items()))})
|
||||
return rows
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
|
||||
parser.add_argument("--duration", type=float, default=20)
|
||||
parser.add_argument("--workers", type=int, default=4)
|
||||
args = parser.parse_args()
|
||||
load = Load(args.workers).start()
|
||||
time.sleep(args.duration)
|
||||
load.stop()
|
||||
print(json.dumps({"total": load.window(0, load.now()), "history": load.verify_history(),
|
||||
"client_retries_same_key": load.retries}, ensure_ascii=False, indent=2))
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1,75 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Level 4: scripts/rolling-update.sh while bench/loadgen.py traffic runs through the LB.
|
||||
|
||||
python bench/rolling_under_load.py --output evidence/rolling-update
|
||||
|
||||
Writes <output>/summary.json (errors, RPS, history check, revisions before/after),
|
||||
<output>/timeline.csv (per second ok/failed/instances) and <output>/script-output.txt.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import csv
|
||||
import json
|
||||
import os
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
sys.path.insert(0, str(ROOT / "scripts"))
|
||||
sys.path.insert(0, str(ROOT / "bench"))
|
||||
import lbctl # noqa: E402
|
||||
from loadgen import Load # noqa: E402
|
||||
|
||||
|
||||
def revisions() -> dict[str, str]:
|
||||
result = {}
|
||||
for r in lbctl.replicas():
|
||||
if r["state"] == "running":
|
||||
label = subprocess.run(["docker", "inspect", "-f", '{{index .Config.Labels "org.opencontainers.image.revision"}}',
|
||||
r["id"]], capture_output=True, text=True).stdout.strip()
|
||||
result[r["name"]] = f"{r['hostname']} revision={label}"
|
||||
return result
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
|
||||
parser.add_argument("--output", default="evidence/rolling-update")
|
||||
parser.add_argument("--workers", type=int, default=4)
|
||||
args = parser.parse_args()
|
||||
out = ROOT / args.output
|
||||
out.mkdir(parents=True, exist_ok=True)
|
||||
before = revisions()
|
||||
load = Load(args.workers).start()
|
||||
time.sleep(5)
|
||||
t_begin = load.now()
|
||||
bash = (os.path.join(os.environ.get("ProgramFiles", r"C:\Program Files"), "Git", "bin", "bash.exe")
|
||||
if os.name == "nt" else "sh")
|
||||
env = {**os.environ, "PYTHON": sys.executable}
|
||||
script = subprocess.run([bash, str(ROOT / "scripts" / "rolling-update.sh")], cwd=ROOT, env=env,
|
||||
capture_output=True, text=True, encoding="utf-8", timeout=900)
|
||||
t_end = load.now()
|
||||
time.sleep(5)
|
||||
load.stop()
|
||||
(out / "script-output.txt").write_text(f"$ sh scripts/rolling-update.sh (exit {script.returncode})\n"
|
||||
+ script.stdout + script.stderr, encoding="utf-8")
|
||||
with (out / "timeline.csv").open("w", newline="", encoding="utf-8") as f:
|
||||
writer = csv.DictWriter(f, fieldnames=["second", "ok", "failed", "instances"])
|
||||
writer.writeheader()
|
||||
writer.writerows(load.per_second())
|
||||
summary = {"script_exit_code": script.returncode, "update_seconds": round(t_end - t_begin, 1),
|
||||
"before": before, "after": revisions(),
|
||||
"during_update": load.window(t_begin, t_end), "whole_run": load.window(0, load.now()),
|
||||
"client_retries_same_key": load.retries, "history": load.verify_history(),
|
||||
"workload": f"{args.workers} closed-loop workers: POST (Idempotency-Key) + GET page via 127.0.0.1:8080"}
|
||||
(out / "summary.json").write_text(json.dumps(summary, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
print(json.dumps({k: summary[k] for k in ("script_exit_code", "update_seconds", "before", "after", "history")},
|
||||
ensure_ascii=False, indent=2))
|
||||
print(json.dumps(summary["during_update"], ensure_ascii=False))
|
||||
return 0 if script.returncode == 0 else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
+184
@@ -0,0 +1,184 @@
|
||||
# Lab 3 stand: client -> HAProxy (lb) -> API x N (one scalable service `api`) -> PostgreSQL + Valkey.
|
||||
# From a clean clone (details and Windows notes: REPORT.md, docs/WINDOWS.md):
|
||||
# python scripts/init_secrets.py # once: random local credentials into ./secrets (git-ignored)
|
||||
# cp .env.example .env # optional overrides (API_REPLICAS, limits, ...)
|
||||
# docker compose --env-file .env up -d --build --wait
|
||||
# Order: postgres/valkey healthy -> migrate (ONE run, exits 0) -> api replicas healthy -> lb.
|
||||
# Images are pinned by version AND digest. Passwords are passed only as secret files, and
|
||||
# every service gets only the secrets it needs. Only the LB publishes ports (loopback only).
|
||||
# Two networks: `edge` (lb <-> api) and `data` (api/migrate <-> postgres/valkey, internal):
|
||||
# the LB cannot reach the stores, and the stores are not reachable from the edge.
|
||||
name: mtusi-lab3
|
||||
|
||||
x-api-env: &api-env
|
||||
APP_ENV: ${APP_ENV:-lab}
|
||||
LOG_LEVEL: ${LOG_LEVEL:-info}
|
||||
SESSION_TTL_SECONDS: ${SESSION_TTL_SECONDS:-3600}
|
||||
SESSION_STORE: ${SESSION_STORE:-valkey}
|
||||
DATABASE_URL_FILE: /run/secrets/database_url
|
||||
CACHE_URL_FILE: /run/secrets/cache_url
|
||||
DB_POOL_MIN: ${DB_POOL_MIN:-1}
|
||||
DB_POOL_MAX: ${DB_POOL_MAX:-10}
|
||||
DB_TIMEOUT_SECONDS: ${DB_TIMEOUT_SECONDS:-3}
|
||||
CACHE_TIMEOUT_SECONDS: ${CACHE_TIMEOUT_SECONDS:-0.5}
|
||||
PASSWORD_HASH_CONCURRENCY: ${PASSWORD_HASH_CONCURRENCY:-2}
|
||||
IDEMPOTENCY_TTL_SECONDS: ${IDEMPOTENCY_TTL_SECONDS:-86400}
|
||||
# Graceful stop: 3 s drain (readiness 503, still serving) + up to 8 s for in-flight requests.
|
||||
DRAIN_SECONDS: ${DRAIN_SECONDS:-3}
|
||||
SHUTDOWN_TIMEOUT_SECONDS: ${SHUTDOWN_TIMEOUT_SECONDS:-8}
|
||||
DOCS_ENABLED: ${DOCS_ENABLED:-true}
|
||||
# Track 1 improvement: `function` = one SQL call per send (default), `multi` = old path (A/B).
|
||||
MESSAGE_WRITE_PATH: ${MESSAGE_WRITE_PATH:-function}
|
||||
|
||||
services:
|
||||
postgres:
|
||||
image: postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c
|
||||
environment:
|
||||
POSTGRES_USER: messenger
|
||||
POSTGRES_DB: messenger
|
||||
POSTGRES_PASSWORD_FILE: /run/secrets/postgres_password
|
||||
# pg_stat_statements: per-statement counters for the track 5 experiments (bench/track1).
|
||||
command: ["postgres", "-c", "shared_preload_libraries=pg_stat_statements", "-c", "max_connections=${PG_MAX_CONNECTIONS:-100}"]
|
||||
secrets: [postgres_password]
|
||||
volumes:
|
||||
- pgdata:/var/lib/postgresql # PostgreSQL 18 image keeps PGDATA under this path
|
||||
healthcheck:
|
||||
test: ["CMD-SHELL", "pg_isready -U messenger -d messenger -q"]
|
||||
interval: 2s
|
||||
timeout: 3s
|
||||
retries: 30
|
||||
start_period: 10s
|
||||
stop_grace_period: 30s # clean shutdown checkpoint instead of crash recovery
|
||||
restart: unless-stopped
|
||||
networks: [data]
|
||||
deploy:
|
||||
resources:
|
||||
limits: {cpus: "${PG_CPUS:-2.0}", memory: "${PG_MEMORY:-1g}"}
|
||||
|
||||
valkey:
|
||||
image: valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571
|
||||
# The password goes into a config include readable only by the valkey user,
|
||||
# so it never appears in the process arguments. docker-entrypoint.sh drops root.
|
||||
command:
|
||||
- sh
|
||||
- -c
|
||||
- >-
|
||||
umask 077 &&
|
||||
printf 'requirepass %s\n' "$$(cat /run/secrets/valkey_password)" > /tmp/requirepass.conf &&
|
||||
chown valkey /tmp/requirepass.conf &&
|
||||
exec docker-entrypoint.sh valkey-server /etc/valkey/valkey.conf
|
||||
secrets: [valkey_password]
|
||||
volumes:
|
||||
- ./infra/valkey/valkey.conf:/etc/valkey/valkey.conf:ro
|
||||
- valkeydata:/data
|
||||
healthcheck:
|
||||
test: ["CMD-SHELL", "REDISCLI_AUTH=\"$$(cat /run/secrets/valkey_password)\" valkey-cli ping | grep -q PONG"]
|
||||
interval: 2s
|
||||
timeout: 3s
|
||||
retries: 30
|
||||
start_period: 5s
|
||||
restart: unless-stopped
|
||||
networks: [data]
|
||||
deploy:
|
||||
resources:
|
||||
limits: {cpus: "${VALKEY_CPUS:-0.5}", memory: 256m}
|
||||
|
||||
# One-shot schema migration: runs ONCE per `up`, before any API replica starts
|
||||
# (api depends_on service_completed_successfully). Replicas never migrate themselves.
|
||||
migrate:
|
||||
build:
|
||||
context: .
|
||||
args:
|
||||
BUILD_REVISION: ${BUILD_REVISION:-dev}
|
||||
image: ${API_IMAGE:-mtusi-messenger:lab3}
|
||||
command: ["python", "-m", "messenger.migrate"]
|
||||
environment:
|
||||
DATABASE_URL_FILE: /run/secrets/database_url
|
||||
secrets: [database_url]
|
||||
depends_on:
|
||||
postgres:
|
||||
condition: service_healthy
|
||||
healthcheck:
|
||||
disable: true
|
||||
restart: "no"
|
||||
networks: [data]
|
||||
deploy:
|
||||
resources:
|
||||
limits: {cpus: "0.5", memory: 256m}
|
||||
|
||||
api:
|
||||
build:
|
||||
context: .
|
||||
args:
|
||||
BUILD_REVISION: ${BUILD_REVISION:-dev}
|
||||
image: ${API_IMAGE:-mtusi-messenger:lab3}
|
||||
environment: *api-env
|
||||
secrets: [database_url, cache_url]
|
||||
# No `ports` and no container_name: replicas are reachable only through `lb`,
|
||||
# and `--scale api=N` / deploy.replicas works. X-Instance-Id = container hostname.
|
||||
expose: ["8080"]
|
||||
healthcheck:
|
||||
test: ["CMD", "python", "-m", "messenger.healthcheck"]
|
||||
interval: 2s
|
||||
timeout: 2s
|
||||
retries: 3
|
||||
start_period: 5s
|
||||
depends_on:
|
||||
postgres:
|
||||
condition: service_healthy
|
||||
valkey:
|
||||
condition: service_healthy
|
||||
migrate:
|
||||
condition: service_completed_successfully
|
||||
stop_grace_period: 15s # > DRAIN_SECONDS + SHUTDOWN_TIMEOUT_SECONDS (11 s)
|
||||
restart: unless-stopped
|
||||
networks: [edge, data]
|
||||
deploy:
|
||||
replicas: ${API_REPLICAS:-2}
|
||||
resources:
|
||||
limits: {cpus: "${API_CPUS:-1.0}", memory: "${API_MEMORY:-256m}"}
|
||||
|
||||
lb:
|
||||
image: haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122
|
||||
volumes:
|
||||
- ./infra/haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro
|
||||
ports:
|
||||
# The ONLY public entry point, on the loopback interfaces only. Both IPv4 and IPv6:
|
||||
# on Windows `localhost` resolves to ::1 first, and a refused ::1 connection costs ~2 s
|
||||
# per request before the fallback to 127.0.0.1 (measured: 2.0 s vs 0.007 s).
|
||||
- "127.0.0.1:${APP_PUBLISH_PORT:-8080}:8080"
|
||||
- "[::1]:${APP_PUBLISH_PORT:-8080}:8080"
|
||||
- "127.0.0.1:${LB_STATS_PORT:-8404}:8404" # local status page (UP/DOWN of replicas)
|
||||
healthcheck:
|
||||
# Config is valid and the process answers; whether any API is ready is the LB's own job.
|
||||
test: ["CMD-SHELL", "haproxy -c -q -f /usr/local/etc/haproxy/haproxy.cfg && wget -q -O /dev/null http://127.0.0.1:8404/"]
|
||||
interval: 5s
|
||||
timeout: 3s
|
||||
retries: 3
|
||||
depends_on:
|
||||
api:
|
||||
condition: service_healthy
|
||||
restart: unless-stopped
|
||||
networks: [edge]
|
||||
deploy:
|
||||
resources:
|
||||
limits: {cpus: "${LB_CPUS:-1.0}", memory: 128m}
|
||||
|
||||
networks:
|
||||
edge:
|
||||
data:
|
||||
internal: true # no route outside: stores talk only to api/migrate
|
||||
|
||||
volumes:
|
||||
pgdata:
|
||||
valkeydata:
|
||||
|
||||
secrets:
|
||||
postgres_password:
|
||||
file: ./secrets/postgres_password
|
||||
valkey_password:
|
||||
file: ./secrets/valkey_password
|
||||
database_url:
|
||||
file: ./secrets/database_url
|
||||
cache_url:
|
||||
file: ./secrets/cache_url
|
||||
@@ -0,0 +1,43 @@
|
||||
# Модель данных lab3
|
||||
|
||||
Схема: [0001_init.sql](../migrations/0001_init.sql) (lab2), [0002_idempotency_keys.sql](../migrations/0002_idempotency_keys.sql) и [0003_post_message_function.sql](../migrations/0003_post_message_function.sql) (lab3). Миграции применяет один раз отдельный сервис `migrate` до старта реплик API; обе миграции lab3 только добавляют объекты, поэтому совместимы с работающей предыдущей версией (rolling update). Источник истины для пользователей, чатов, участников и сообщений — PostgreSQL 18. Для сессий источник истины — хранилище, выбранное через `SESSION_STORE`. По умолчанию это Valkey с AOF, его можно переключить на таблицу `sessions` в PostgreSQL.
|
||||
|
||||
## Таблицы и ограничения
|
||||
|
||||
| Таблица | Ключ | Ограничения | Зачем |
|
||||
| --- | --- | --- | --- |
|
||||
| `users` | `id uuid` | `UNIQUE (username)`, `CHECK` формата username, длина `display_name` | уникальность имени держит БД, а не код: гонку регистрации разрешает уникальный индекс |
|
||||
| `chats` | `id uuid` | FK `created_by → users`, длина `title` | `last_message_at` — «часы» чата для упорядочивания сообщений |
|
||||
| `chat_members` | `(chat_id, user_id)` | FK на `chats` и `users`, `UNIQUE (chat_id, position)` | `position` хранит порядок `member_ids`; `chat_created_at` — копия времени чата для keyset-списка чатов |
|
||||
| `messages` | `id uuid` | FK `(chat_id, sender_id) → chat_members`, длина `text` | отправитель обязан быть участником — это гарантирует сама БД |
|
||||
| `sessions` | `token_hash bytea` | FK `user_id → users`, длина хеша 32 байта | хранится SHA-256 токена; сам токен не хранится нигде |
|
||||
| `idempotency_keys` | `(user_id, chat_id, key)` | FK на `users`, `chats`; `message_id → messages` (DEFERRABLE, `UNIQUE`); `CHECK` формата ключа и длины SHA-256 | общая для всех реплик идемпотентность отправки: ключ → отпечаток тела → сообщение. Хранится ≥ `IDEMPOTENCY_TTL_SECONDS` (24 ч), очистка — пачками раз в 10 мин под advisory-замком |
|
||||
| `schema_migrations` | `version` | checksum файла | какие миграции применены; изменённую задним числом миграцию мигратор не применит |
|
||||
|
||||
**Почему `bytea` для текста.** Контракт разрешает символ U+0000, а тип `text` в PostgreSQL не может содержать байт NUL (`DataError`). Поэтому `display_name`, `title` и `text` хранятся как байты UTF-8. Длину в code points проверяет API, а `CHECK` в байтах — страховка (1 code point ≤ 4 байта).
|
||||
|
||||
## Индексы и запросы
|
||||
|
||||
- `messages_chat_keyset (chat_id, created_at, id)`. Первая страница — `WHERE chat_id=$1 ORDER BY created_at, id LIMIT n`, следующая — `WHERE chat_id=$1 AND (created_at, id) > ($2, $3) ORDER BY ... LIMIT n`. Сравнение строк `(created_at, id) > (...)` попадает в `Index Cond`: поиск начинается прямо с позиции cursor и читает только `n+1` строк. Запросов два, а не один с `($2 IS NULL OR ...)`: иначе при generic plan условие cursor уходит в `Filter`, и поздняя страница сканирует всю историю чата до неё.
|
||||
- `chat_members_user_keyset (user_id, chat_created_at, chat_id)` — список чатов пользователя тем же keyset-способом.
|
||||
- `users_username_key` — вход по username и уникальность.
|
||||
- `sessions_expires_at` — очистка истёкших сессий.
|
||||
|
||||
## Транзакционные границы
|
||||
|
||||
| Операция | Транзакция | Что гарантирует |
|
||||
| --- | --- | --- |
|
||||
| Регистрация | один `INSERT` | `UniqueViolation` → 409; ровно одна строка при гонке |
|
||||
| Создание чата | `INSERT chats` → `INSERT chat_members` (все участники одним запросом) → `COMMIT` | неизвестный участник вызывает FK-ошибку **после** вставки чата; откат удаляет и чат, и частичное членство |
|
||||
| Сообщение (lab3, с ключом) | проверка членства → `INSERT idempotency_keys … ON CONFLICT DO UPDATE … WHERE ключ старше 24 ч` → если вставлен: `UPDATE chats … RETURNING` → `INSERT messages`; если нет: прочитать сохранённое сообщение и сравнить отпечаток → `COMMIT`. По умолчанию всё это — одна SQL-функция `post_message` в autocommit (одно обращение к БД) | ключ, сообщение и «результат» коммитятся вместе или не коммитятся вовсе: «висящего» ключа нет. Конкурентный повтор ждёт на уникальном индексе завершения первой транзакции и затем видит её результат (READ COMMITTED, новый снимок на каждый оператор). Другое тело → 409, тот же — тот же Message |
|
||||
| Сообщение (lab2-часть) | проверка членства → `UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + 1 мкс) RETURNING` → `INSERT messages` → `COMMIT` | строка чата заблокирована до конца транзакции, поэтому писатели одного чата идут по очереди. Время сообщений строго растёт **в порядке коммитов**: страница по cursor не пропустит сообщение, закоммиченное «задним числом» |
|
||||
| Вход | `SELECT` пользователя, проверка Argon2id, запись сессии | неверный пароль ничего не пишет |
|
||||
| Logout | `DELETE` сессии (или `DEL` в Valkey) | повтор с тем же токеном — 401 |
|
||||
|
||||
Чтение истории не требует snapshot между HTTP-запросами (так в контракте): новое сообщение появится на следующей странице. Список чатов сортируется по `chats.created_at = clock_timestamp()` без блокировки. Два чата, создаваемые одновременно, могут закоммититься не в порядке времени. Тогда у обхода, который уже начался, такой чат может оказаться «до» cursor; новый обход его увидит. Чаты создаются редко, поэтому мы не сериализуем их глобальным замком; это ограничение описано в REPORT.md.
|
||||
|
||||
## Сессии
|
||||
|
||||
- Токен: `secrets.token_urlsafe(32)` — 256 бит случайности, в ответе 43 символа. В хранилище лежит только SHA-256 токена. Для случайного 256-битного значения быстрого хеша достаточно, медленный KDF нужен только для низкоэнтропийных паролей.
|
||||
- `SESSION_STORE=valkey` (по умолчанию): ключ `session:<sha256-hex>`, значение `{user_id, expires_at}`. Срок жизни задаётся абсолютным `PXAT expires_at`, поэтому рестарт его не продлевает. Включён `appendonly yes` + `appendfsync always`: подтверждённые вход и logout сохраняются до ответа клиенту. Valkey — **единственное** хранилище сессий, это не кэш над PostgreSQL. Поэтому нет второй копии, которая могла бы «оживить» отозванный токен после отказа. Если Valkey недоступен, проверить сессию нельзя: защищённые запросы и logout получают 503, анонимный доступ не открывается.
|
||||
- `SESSION_STORE=postgres`: таблица `sessions`, проверка `expires_at > now()` в запросе.
|
||||
+138
@@ -0,0 +1,138 @@
|
||||
# Запуск на ноутбуке с Windows и сценарий защиты ЛР 3
|
||||
|
||||
Команды — в **PowerShell** из папки проекта (где лежит `compose.yaml`). Переменные `$env:...` действуют до закрытия окна.
|
||||
|
||||
## 1. Что установить один раз
|
||||
|
||||
| Программа | Зачем | Важно |
|
||||
| --- | --- | --- |
|
||||
| [Git for Windows](https://git-scm.com/download/win) | clone и Git Bash | путь по умолчанию `C:\Program Files\Git`: оттуда `scripts\*.cmd` запускают `.sh`-хуки |
|
||||
| [Docker Desktop](https://www.docker.com/products/docker-desktop/) | PostgreSQL, Valkey, HAProxy, API | backend WSL 2, виртуализация в BIOS; для трека 1 в Settings → Resources дать **8 CPU** |
|
||||
| [Python 3.12+](https://www.python.org/downloads/windows/) | secrets, публичные и свои тесты, скрипты замеров | галочка **Add python.exe to PATH** |
|
||||
| GNU make (необязательно) | `make check/test/test-failover …` | `winget install ezwinports.make`, затем **новое** окно PowerShell. Рецептам Makefile нужен `sh`: в каждом окне PowerShell один раз `$env:Path += ";C:\Program Files\Git\usr\bin"` (или запускать make из Git Bash). Без make те же команды есть в таблице REPORT.md |
|
||||
|
||||
Проверка:
|
||||
|
||||
```powershell
|
||||
git --version; docker version --format '{{.Server.Version}}'; docker compose version; python --version; make --version
|
||||
```
|
||||
|
||||
## 2. Развернуть проект (дома, заранее — нужен интернет)
|
||||
|
||||
```powershell
|
||||
git clone https://gitea.privcloud.tech/wfa/mtusi2026lab3 mtusi2026lab3
|
||||
cd mtusi2026lab3
|
||||
git checkout solution/bvt2302-bodanyuk
|
||||
python scripts\init_secrets.py # случайные пароли в .\secrets (не в Git)
|
||||
Copy-Item .env.example .env
|
||||
docker compose --env-file .env config --quiet # конфигурация собирается без ошибок
|
||||
docker compose --env-file .env up -d --build --wait
|
||||
```
|
||||
|
||||
Первый запуск качает образы (~800 МБ) и собирает API — 3–10 минут. Команда завершается, когда postgres, valkey, обе реплики api и lb в состоянии `healthy`, а `migrate` завершился с кодом 0.
|
||||
|
||||
Окружение для своих тестов и скриптов (один раз):
|
||||
|
||||
```powershell
|
||||
python -m venv .venv
|
||||
.venv\Scripts\python -m pip install --require-hashes --no-deps -r requirements.txt
|
||||
.venv\Scripts\python -m pip install -r requirements-test.txt -r requirements-dev.txt
|
||||
$env:PYTHONUTF8 = "1" # публичные тесты читают UTF-8 JSON
|
||||
```
|
||||
|
||||
Проверка в браузере: http://127.0.0.1:8080/health/ready → `{"status":"ok"}`; Swagger — http://127.0.0.1:8080/docs; состояние балансировщика (какие реплики UP/DOWN) — http://127.0.0.1:8404/.
|
||||
|
||||
## 3. Перед защитой (в аудитории)
|
||||
|
||||
```powershell
|
||||
cd mtusi2026lab3
|
||||
$env:PYTHONUTF8 = "1"
|
||||
$env:Path += ";C:\Program Files\Git\usr\bin" # sh для рецептов Makefile
|
||||
docker compose up -d --wait # если стенд был остановлен
|
||||
make check test PYTHON=python # или: python scripts\check_contract.py; python tests\smoke.py
|
||||
python scripts\lbctl.py status # обе реплики UP
|
||||
```
|
||||
|
||||
## 4. Сценарий защиты из README — по шагам
|
||||
|
||||
**Шаг 1. ≥ 2 разных `X-Instance-Id` через LB с одной сессией.**
|
||||
|
||||
```powershell
|
||||
python scripts\lbctl.py replicas # контейнеры api: имя, X-Instance-Id (= hostname), PID, IP
|
||||
# Один пользователь, один токен, 6 запросов подряд через LB — номера реплик чередуются:
|
||||
$u = "demo_" + (Get-Random); $p = "Demo-password-123"
|
||||
Invoke-RestMethod -Method Post http://127.0.0.1:8080/api/v1/users -ContentType 'application/json' -Body (@{username=$u; display_name='Демо'; password=$p} | ConvertTo-Json)
|
||||
$basic = [Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes("${u}:${p}"))
|
||||
$t = (Invoke-RestMethod -Method Post http://127.0.0.1:8080/api/v1/auth/sessions -Headers @{Authorization="Basic $basic"}).token
|
||||
1..6 | ForEach-Object { (Invoke-WebRequest http://127.0.0.1:8080/api/v1/users/me -Headers @{Authorization="Bearer $t"} -UseBasicParsing).Headers['X-Instance-Id'] }
|
||||
```
|
||||
|
||||
Повторный вход не нужен: сессия лежит в общем Valkey. То же автоматически: `.venv\Scripts\python -m pytest student-tests\test_balancing.py -k one_session -v`.
|
||||
|
||||
**Шаг 2. Отказ реплики публичным тестом.**
|
||||
|
||||
```powershell
|
||||
make test-failover ACTION_SCRIPT=scripts/stop-one.cmd PYTHON=python
|
||||
# без make: python tests\resilience.py --mode failover --action-script scripts\stop-one.cmd
|
||||
```
|
||||
|
||||
Тест сам выбирает реплику, обслужившую его первое сообщение, и передаёт её `X-Instance-Id` хуку. Хук останавливает именно этот контейнер и оставляет его остановленным. Вариант «падение процесса» — `$env:STOP_MODE = "kill"` перед командой (SIGKILL), вернуть — `Remove-Item Env:STOP_MODE`.
|
||||
|
||||
**Шаг 3. Показать, какой контейнер действительно остановлен, и повторить для второй.**
|
||||
|
||||
```powershell
|
||||
python scripts\lbctl.py replicas # у остановленной: exited, exit_code 0 (stop) или 137 (kill)
|
||||
python scripts\lbctl.py status # её слот в HAProxy — MAINT/DOWN, вторая UP
|
||||
docker compose logs --tail 20 api # drain_started -> shutdown_requested -> stopped у остановленной
|
||||
make restore PYTHON=python # или: & "C:\Program Files\Git\bin\bash.exe" scripts/restore-api.sh; python scripts\lbctl.py wait-up 2
|
||||
make test-failover ACTION_SCRIPT=scripts/stop-one.cmd PYTHON=python
|
||||
```
|
||||
|
||||
Тест останавливает ту реплику, которая обслужила его первое сообщение, поэтому для второй он может снова выбрать прежнюю. Детерминированно по очереди каждую реплику останавливает серия уровня 4 (~5 мин): `python bench\failover_series.py --output evidence\failover` (stop, kill и pause для каждой реплики, с восстановлением предыдущей).
|
||||
|
||||
**Шаг 4. Restart всего стенда с сохранением volumes и прежним токеном.**
|
||||
|
||||
```powershell
|
||||
make test-persistence ACTION_SCRIPT=scripts/restart.cmd PYTHON=python
|
||||
python scripts\lbctl.py replicas # у всех api свежий started и новый PID; LB снова видит обе реплики
|
||||
docker compose ps # postgres, valkey, lb — тоже «Up N seconds»
|
||||
```
|
||||
|
||||
**Уровень 4 вживую.**
|
||||
|
||||
```powershell
|
||||
# Неготовая, но живая реплика исключается из LB (pg_hba reject и docker network disconnect), ~1 мин:
|
||||
.venv\Scripts\python -m pytest student-tests\test_balancing.py -k not_ready -v -s
|
||||
# Rolling update под нагрузкой (~1,5 мин): новая ревизия на обеих репликах, 0 ошибок
|
||||
python bench\rolling_under_load.py --output evidence\rolling-update
|
||||
# Одновременные повторы с одним ключом через разные реплики:
|
||||
.venv\Scripts\python -m pytest student-tests\test_idempotency.py -k concurrent -v
|
||||
```
|
||||
|
||||
**Трек 1.** Полная серия идёт часы, поэтому на защите — готовые результаты: [evidence/track1/summary.md](../evidence/track1/summary.md) и графики `evidence/track1/*/*/throughput.svg`. Вживую можно показать профиль одной реплики (~2 мин) и одну короткую серию (~6 мин):
|
||||
|
||||
```powershell
|
||||
docker compose stop # основной стенд не должен мешать замеру
|
||||
python bench\track1\profile.py --label demo --rate 240
|
||||
python bench\track1\run.py --series demo --label now --rates 150,250,350 --repeats 1 --config replicas=2,MESSAGE_WRITE_PATH=multi --config replicas=2,MESSAGE_WRITE_PATH=function
|
||||
docker compose up -d --wait # вернуть основной стенд
|
||||
```
|
||||
|
||||
**Все свои тесты** (~8 мин): `.venv\Scripts\python -m pytest student-tests -v`.
|
||||
|
||||
## 5. Если что-то пошло не так
|
||||
|
||||
| Симптом | Что сделать |
|
||||
| --- | --- |
|
||||
| `error during connect … dockerDesktopLinuxEngine` | Docker Desktop не запущен — запустить и подождать минуту |
|
||||
| `port is already allocated` на 8080/8404 | занято другим стендом (lab2: `docker compose -p mtusi-lab2 stop`) или в `.env` поставить `APP_PUBLISH_PORT=8081`, `LB_STATS_PORT=8405`; тестам — `--base-url http://127.0.0.1:8081` / `$env:BASE_URL` |
|
||||
| `secrets/... not found` при `up` | не выполнен `python scripts\init_secrets.py` |
|
||||
| `UnicodeDecodeError` в `tests\*.py` | не задан `$env:PYTHONUTF8 = "1"` |
|
||||
| `*.cmd`: «не удаётся найти путь» | Git установлен не в `C:\Program Files\Git` — поправить путь в `scripts\*.cmd` |
|
||||
| `make: command not found` | новое окно PowerShell после установки make или команды без make из REPORT.md |
|
||||
| make: `'test' is not recognized` | в этом окне не добавлен `C:\Program Files\Git\usr\bin` в `$env:Path` (рецептам нужен `sh`) |
|
||||
| после `stop-one` в LB одна реплика | так и задумано; вернуть: `make restore PYTHON=python` |
|
||||
| api `unhealthy` после сна ноутбука | `docker compose restart` и `python scripts\lbctl.py wait-up 2` |
|
||||
| нужно начать с чистого листа | `docker compose down -v` (удаляет учебные данные) → `docker compose up -d --build --wait` |
|
||||
|
||||
Остановить стенд после защиты: `docker compose down` (данные сохранятся) или `docker compose down -v` (данные удаляются).
|
||||
@@ -0,0 +1,91 @@
|
||||
# HAProxy 3.2 LTS — the single public entry point of the lab 3 stand (compose service `lb`).
|
||||
#
|
||||
# Backend discovery: Docker DNS name `api` returns one A record per RUNNING replica of the
|
||||
# compose service `api` (deploy.replicas / --scale). server-template keeps 8 slots and
|
||||
# re-resolves every second, so scale up/down, a restarted container with a new IP and a
|
||||
# full stack restart are picked up without restarting the LB. A stopped container drops
|
||||
# out of DNS and its slot goes to MAINT.
|
||||
#
|
||||
# Readiness: an ACTIVE check `GET /health/ready` (PostgreSQL + Valkey reachable, not draining)
|
||||
# every 1 s; 2 failures -> DOWN, 2 successes -> UP. Passive errors do not decide anything on
|
||||
# their own, and there are NO backup/fallback servers without checks: a replica that answers
|
||||
# /health/live but not /health/ready never gets business traffic.
|
||||
# Worst case detection of a failed replica (a check answered at once counts at once, a check
|
||||
# that hangs counts only after `timeout check`):
|
||||
# fast explicit 503 (drain after SIGTERM): inter 1 s + fastinter 0.5 s ~= 1.5 s;
|
||||
# slow ready (PostgreSQL/Valkey unreachable: the answer takes ~1 s = timeout check)
|
||||
# or frozen process: inter 1 s + 1 s + fastinter 0.5 s + 1 s ~= 3.5 s.
|
||||
# Requests that could not even be SENT (connection refused/reset before the request)
|
||||
# are retried on another replica (retry-on conn-failure + redispatch) - safe for POST,
|
||||
# because the API has not seen them. Nothing that may have reached the API is retried by
|
||||
# the LB: an uncertain POST is retried by the CLIENT with the same Idempotency-Key.
|
||||
# No sticky sessions: sessions live in Valkey, every replica serves every user.
|
||||
|
||||
global
|
||||
log stdout format raw local0 info
|
||||
maxconn 4096
|
||||
# Runtime API for `show servers state` / `show stat` (scripts and tests, no network port).
|
||||
stats socket /var/lib/haproxy/admin.sock mode 600 level admin
|
||||
hard-stop-after 15s
|
||||
|
||||
resolvers docker
|
||||
nameserver dns1 127.0.0.11:53
|
||||
accepted_payload_size 8192
|
||||
resolve_retries 3
|
||||
timeout resolve 1s
|
||||
timeout retry 1s
|
||||
hold valid 1s
|
||||
hold other 1s
|
||||
hold refused 1s
|
||||
hold nx 1s
|
||||
hold timeout 1s
|
||||
hold obsolete 1s
|
||||
|
||||
defaults
|
||||
mode http
|
||||
log global
|
||||
option httplog
|
||||
option dontlognull
|
||||
option http-keep-alive
|
||||
timeout connect 1s
|
||||
timeout client 30s
|
||||
timeout server 10s # > DB_TIMEOUT_SECONDS (3 s): the API answers 503 itself first
|
||||
timeout http-request 10s
|
||||
timeout http-keep-alive 10s
|
||||
timeout queue 3s
|
||||
timeout check 1s
|
||||
retries 2
|
||||
retry-on conn-failure
|
||||
option redispatch 1
|
||||
# Errors generated by the LB itself, in the contract error format with the same request id.
|
||||
# They carry no X-Instance-Id: no API replica produced them. 502/504 = the outcome of the
|
||||
# request is unknown (a replica died mid-request / did not answer in time): retry with the
|
||||
# same Idempotency-Key. 503 = no ready replica at all.
|
||||
http-error status 502 content-type application/json lf-string "{\"error\":{\"code\":\"unavailable\",\"message\":\"API-реплика оборвала ответ; повторите с тем же Idempotency-Key\",\"request_id\":\"%[var(txn.request_id)]\"}}" hdr X-Request-Id "%[var(txn.request_id)]"
|
||||
http-error status 503 content-type application/json lf-string "{\"error\":{\"code\":\"unavailable\",\"message\":\"Нет готовых API-реплик\",\"request_id\":\"%[var(txn.request_id)]\"}}" hdr X-Request-Id "%[var(txn.request_id)]"
|
||||
http-error status 504 content-type application/json lf-string "{\"error\":{\"code\":\"unavailable\",\"message\":\"API-реплика не ответила вовремя; повторите с тем же Idempotency-Key\",\"request_id\":\"%[var(txn.request_id)]\"}}" hdr X-Request-Id "%[var(txn.request_id)]"
|
||||
|
||||
frontend public
|
||||
bind :8080
|
||||
# One request id end to end: keep a valid client X-Request-Id, otherwise create one; the
|
||||
# API echoes it, and errors generated by the LB itself carry the same value.
|
||||
http-request set-header X-Request-Id %[uuid()] unless { req.hdr(x-request-id) -m reg ^[A-Za-z0-9._:-]{1,128}$ }
|
||||
http-request set-var(txn.request_id) req.hdr(x-request-id)
|
||||
log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq rid=%[var(txn.request_id)] \"%HM %HPO\""
|
||||
default_backend api
|
||||
|
||||
backend api
|
||||
balance roundrobin
|
||||
option httpchk
|
||||
http-check send meth GET uri /health/ready ver HTTP/1.1 hdr Host api
|
||||
http-check expect status 200
|
||||
default-server inter 1s fastinter 500ms downinter 1s fall 2 rise 2 check resolvers docker init-addr none resolve-prefer ipv4 maxconn 200
|
||||
server-template api 8 api:8080
|
||||
|
||||
# Local-only status page (127.0.0.1:8404 on the host): which replica is UP/DOWN right now.
|
||||
frontend stats
|
||||
bind :8404
|
||||
stats enable
|
||||
stats uri /
|
||||
stats refresh 2s
|
||||
stats show-legends
|
||||
@@ -0,0 +1,17 @@
|
||||
# Valkey is the single store of sessions (SESSION_STORE=valkey), so it must be durable.
|
||||
# AOF with fsync on every write: a confirmed login or logout survives a crash/restart,
|
||||
# a logout can never be lost and "resurrect" a token. Sessions are small and rare
|
||||
# writes, so the fsync cost is acceptable here (see REPORT.md).
|
||||
appendonly yes
|
||||
appendfsync always
|
||||
appenddirname "appendonlydir"
|
||||
# RDB snapshots are not needed on top of AOF.
|
||||
save ""
|
||||
dir /data
|
||||
port 6379
|
||||
protected-mode yes
|
||||
# Sessions are never evicted silently: if memory is full, writes fail and login returns 503.
|
||||
maxmemory 128mb
|
||||
maxmemory-policy noeviction
|
||||
# Password, generated by scripts/init_secrets.py and written here at container start.
|
||||
include /tmp/requirepass.conf
|
||||
@@ -0,0 +1,65 @@
|
||||
-- 0001: initial schema of the messenger (lab 2). Design rationale: docs/DATA_MODEL.md
|
||||
-- Applied once by `python -m messenger.migrate`; the version is recorded in schema_migrations.
|
||||
|
||||
-- User-supplied free text (display_name, title, message text) is stored as UTF-8 in BYTEA:
|
||||
-- the contract allows U+0000, and PostgreSQL TEXT cannot contain NUL bytes.
|
||||
-- Lengths are limited in code points by the API; the CHECKs below are byte-level safety nets
|
||||
-- (one code point is at most 4 bytes in UTF-8).
|
||||
|
||||
CREATE TABLE users (
|
||||
id uuid PRIMARY KEY,
|
||||
username text NOT NULL,
|
||||
display_name bytea NOT NULL,
|
||||
password_hash text NOT NULL, -- Argon2id PHC string, never the password
|
||||
created_at timestamptz NOT NULL DEFAULT clock_timestamp(),
|
||||
CONSTRAINT users_username_key UNIQUE (username),
|
||||
CONSTRAINT users_username_format CHECK (username ~ '^[a-z0-9_]{3,32}$'),
|
||||
CONSTRAINT users_display_name_len CHECK (octet_length(display_name) BETWEEN 1 AND 400)
|
||||
);
|
||||
|
||||
CREATE TABLE chats (
|
||||
id uuid PRIMARY KEY,
|
||||
title bytea NOT NULL,
|
||||
created_by uuid NOT NULL REFERENCES users (id),
|
||||
created_at timestamptz NOT NULL DEFAULT clock_timestamp(),
|
||||
-- Per-chat clock: created_at of the newest message. Updating this row serialises writers
|
||||
-- of one chat, so message timestamps strictly increase in commit order (see DATA_MODEL.md).
|
||||
last_message_at timestamptz,
|
||||
CONSTRAINT chats_title_len CHECK (octet_length(title) BETWEEN 1 AND 400)
|
||||
);
|
||||
|
||||
CREATE TABLE chat_members (
|
||||
chat_id uuid NOT NULL REFERENCES chats (id),
|
||||
user_id uuid NOT NULL REFERENCES users (id),
|
||||
position smallint NOT NULL, -- order of member_ids sent by the creator
|
||||
chat_created_at timestamptz NOT NULL, -- copy of chats.created_at for keyset listing
|
||||
PRIMARY KEY (chat_id, user_id),
|
||||
CONSTRAINT chat_members_position_key UNIQUE (chat_id, position)
|
||||
);
|
||||
-- GET /chats of one user: (user_id, created_at, id) range scan, no sort.
|
||||
CREATE INDEX chat_members_user_keyset ON chat_members (user_id, chat_created_at, chat_id);
|
||||
|
||||
CREATE TABLE messages (
|
||||
id uuid PRIMARY KEY,
|
||||
chat_id uuid NOT NULL,
|
||||
sender_id uuid NOT NULL,
|
||||
text bytea NOT NULL,
|
||||
created_at timestamptz NOT NULL,
|
||||
-- The sender must be a member of this chat: enforced by the database, not only by the API.
|
||||
CONSTRAINT messages_sender_is_member FOREIGN KEY (chat_id, sender_id)
|
||||
REFERENCES chat_members (chat_id, user_id),
|
||||
CONSTRAINT messages_text_len CHECK (octet_length(text) BETWEEN 1 AND 16000)
|
||||
);
|
||||
-- History page: WHERE chat_id = $1 AND (created_at, id) > ($2, $3) ORDER BY created_at, id LIMIT n
|
||||
-- becomes an index range scan that starts exactly at the cursor.
|
||||
CREATE INDEX messages_chat_keyset ON messages (chat_id, created_at, id);
|
||||
|
||||
-- Sessions for SESSION_STORE=postgres (the default store is Valkey; see docs/DATA_MODEL.md).
|
||||
CREATE TABLE sessions (
|
||||
token_hash bytea PRIMARY KEY, -- SHA-256 of the token; the token is not stored
|
||||
user_id uuid NOT NULL REFERENCES users (id),
|
||||
created_at timestamptz NOT NULL DEFAULT clock_timestamp(),
|
||||
expires_at timestamptz NOT NULL,
|
||||
CONSTRAINT sessions_token_hash_len CHECK (octet_length(token_hash) = 32)
|
||||
);
|
||||
CREATE INDEX sessions_expires_at ON sessions (expires_at);
|
||||
@@ -0,0 +1,25 @@
|
||||
-- 0002 (lab 3): shared idempotency of POST /chats/{chat_id}/messages.
|
||||
-- Additive only (new table), so the lab 2 API keeps working while this runs: it is safe
|
||||
-- to apply before a rolling update of the API replicas (see REPORT.md, "Миграции").
|
||||
--
|
||||
-- Scope of a key (contracts/README.md): current user + HTTP method + chat path + key.
|
||||
-- The method (POST) and the path template are fixed for this table, so the scope
|
||||
-- is (user_id, chat_id, key). The key row, the message and the result are written in
|
||||
-- ONE transaction (repo.post_message): there is no "pending" state that a crash could
|
||||
-- leave behind, and the stored result is the message row itself.
|
||||
|
||||
CREATE TABLE idempotency_keys (
|
||||
user_id uuid NOT NULL REFERENCES users (id),
|
||||
chat_id uuid NOT NULL REFERENCES chats (id),
|
||||
key text NOT NULL,
|
||||
request_hash bytea NOT NULL, -- SHA-256 of the canonical JSON body
|
||||
-- Inserted before the message in the same transaction, hence checked at COMMIT.
|
||||
message_id uuid NOT NULL REFERENCES messages (id) DEFERRABLE INITIALLY DEFERRED,
|
||||
created_at timestamptz NOT NULL DEFAULT clock_timestamp(),
|
||||
PRIMARY KEY (user_id, chat_id, key),
|
||||
CONSTRAINT idempotency_keys_key_format CHECK (key ~ '^[A-Za-z0-9._:-]{1,128}$'),
|
||||
CONSTRAINT idempotency_keys_hash_len CHECK (octet_length(request_hash) = 32),
|
||||
CONSTRAINT idempotency_keys_message_key UNIQUE (message_id)
|
||||
);
|
||||
-- Housekeeping of keys older than IDEMPOTENCY_TTL_SECONDS (>= 24 h).
|
||||
CREATE INDEX idempotency_keys_created_at ON idempotency_keys (created_at);
|
||||
@@ -0,0 +1,50 @@
|
||||
-- 0003 (lab 3, track 1 improvement): the whole idempotent "send message" as ONE statement.
|
||||
-- Additive (a new function), so replicas of the previous version keep working while it is
|
||||
-- applied, and the previous write path (MESSAGE_WRITE_PATH=multi) stays usable for A/B runs.
|
||||
--
|
||||
-- Same semantics and the same SQL as repo.post_message (multi-statement path), but executed
|
||||
-- inside the server: the API sends one call instead of ~11 round trips in two transactions
|
||||
-- (membership check, BEGIN, key claim, chat clock, message, COMMIT, pool checks). Called in
|
||||
-- autocommit mode, so the function call IS the transaction: key row, chat clock and message
|
||||
-- commit together or not at all. VOLATILE + READ COMMITTED: every statement inside takes a
|
||||
-- fresh snapshot, so after waiting on a concurrent claim of the same key the stored result
|
||||
-- of the winner is visible, exactly as in the multi-statement path.
|
||||
--
|
||||
-- o_status: created | replayed | conflict | not_found | retry (key vanished in between: cleanup)
|
||||
CREATE FUNCTION post_message(p_user uuid, p_chat uuid, p_key text, p_hash bytea, p_message uuid,
|
||||
p_text bytea, p_ttl_seconds integer)
|
||||
RETURNS TABLE (o_status text, o_id uuid, o_text bytea, o_created_at timestamptz)
|
||||
LANGUAGE plpgsql VOLATILE AS $$
|
||||
DECLARE
|
||||
v_created timestamptz;
|
||||
BEGIN
|
||||
PERFORM 1 FROM chat_members m WHERE m.chat_id = p_chat AND m.user_id = p_user;
|
||||
IF NOT FOUND THEN
|
||||
RETURN QUERY SELECT 'not_found'::text, NULL::uuid, NULL::bytea, NULL::timestamptz;
|
||||
RETURN;
|
||||
END IF;
|
||||
|
||||
INSERT INTO idempotency_keys AS k (user_id, chat_id, key, request_hash, message_id)
|
||||
VALUES (p_user, p_chat, p_key, p_hash, p_message)
|
||||
ON CONFLICT (user_id, chat_id, key) DO UPDATE
|
||||
SET request_hash = EXCLUDED.request_hash, message_id = EXCLUDED.message_id, created_at = clock_timestamp()
|
||||
WHERE k.created_at < clock_timestamp() - make_interval(secs => p_ttl_seconds);
|
||||
IF FOUND THEN
|
||||
-- Per-chat clock under the chat row lock (see migrations/0001 and docs/DATA_MODEL.md).
|
||||
UPDATE chats c SET last_message_at = greatest(clock_timestamp(), c.last_message_at + interval '1 microsecond')
|
||||
WHERE c.id = p_chat RETURNING c.last_message_at INTO v_created;
|
||||
INSERT INTO messages (id, chat_id, sender_id, text, created_at)
|
||||
VALUES (p_message, p_chat, p_user, p_text, v_created);
|
||||
RETURN QUERY SELECT 'created'::text, p_message, p_text, v_created;
|
||||
RETURN;
|
||||
END IF;
|
||||
|
||||
RETURN QUERY
|
||||
SELECT CASE WHEN k.request_hash = p_hash THEN 'replayed' ELSE 'conflict' END, m.id, m.text, m.created_at
|
||||
FROM idempotency_keys k JOIN messages m ON m.id = k.message_id
|
||||
WHERE k.user_id = p_user AND k.chat_id = p_chat AND k.key = p_key;
|
||||
IF NOT FOUND THEN
|
||||
RETURN QUERY SELECT 'retry'::text, NULL::uuid, NULL::bytea, NULL::timestamptz;
|
||||
END IF;
|
||||
END
|
||||
$$;
|
||||
@@ -0,0 +1,4 @@
|
||||
# Student tests (in-process, via Starlette TestClient). Runtime deps come from requirements.txt.
|
||||
pytest==9.1.1
|
||||
httpx2==2.13.1
|
||||
jsonschema==4.26.0
|
||||
@@ -0,0 +1,16 @@
|
||||
# Direct runtime dependencies. The pinned lock with hashes is requirements.txt:
|
||||
# docker run --rm -v "$PWD:/w" -w /w python:3.13.16-slim-bookworm \
|
||||
# sh -c "pip install -q pip-tools==7.6.1 && pip-compile --generate-hashes --strip-extras --no-emit-index-url -o requirements.txt requirements.in"
|
||||
starlette==1.7.0
|
||||
uvicorn==0.54.0
|
||||
# Fast C parser/event loop used by uvicorn in the container (Linux only).
|
||||
httptools==0.9.0
|
||||
uvloop==0.23.0 ; sys_platform != "win32"
|
||||
# PostgreSQL driver (libpq bundled in the binary wheel) and its async connection pool.
|
||||
psycopg==3.3.6
|
||||
psycopg-binary==3.3.6
|
||||
psycopg-pool==3.3.3
|
||||
# Valkey client (Valkey speaks the Redis protocol).
|
||||
redis==8.1.0
|
||||
# Argon2id password KDF.
|
||||
argon2-cffi==25.1.0
|
||||
@@ -0,0 +1,452 @@
|
||||
#
|
||||
# This file is autogenerated by pip-compile with Python 3.13
|
||||
# by the following command:
|
||||
#
|
||||
# pip-compile --generate-hashes --no-emit-index-url --no-index --output-file=requirements.txt --strip-extras requirements.in
|
||||
#
|
||||
anyio==4.15.1 \
|
||||
--hash=sha256:6152fdbbf9a77fdec97731721bebf7c4c44f7c29b424b0065826173efc7ed101 \
|
||||
--hash=sha256:9f28306018cbd6d329e64a36d58256edff76dd996fe423bc957326e578b82a94
|
||||
# via starlette
|
||||
argon2-cffi==25.1.0 \
|
||||
--hash=sha256:694ae5cc8a42f4c4e2bf2ca0e64e51e23a040c6a517a85074683d3959e1346c1 \
|
||||
--hash=sha256:fdc8b074db390fccb6eb4a3604ae7231f219aa669a2652e0f20e16ba513d5741
|
||||
# via -r requirements.in
|
||||
argon2-cffi-bindings==26.1.0 \
|
||||
--hash=sha256:061a6919145bbf282ebf1f9c59d3135d4833c25313c8595c0d68cf7712ddfce2 \
|
||||
--hash=sha256:0cc40f7b4050bb93eb67de95d2d759322fc7ce4930b9d645581ecf4913ec651e \
|
||||
--hash=sha256:151dfaad9de753f4af2a7854e707e4784f2acc434340ade64239c5b104b2d605 \
|
||||
--hash=sha256:19423e5d7ac1cc354baab59eaabf18db2ec04ef6593b5abe5a34f323c4a8f87a \
|
||||
--hash=sha256:19b562b1de4b9052ef1214a2821c44b6e6f22945daa102c32ae4eff929d8b6d8 \
|
||||
--hash=sha256:1a0a29ed86960e44eaace7e081bdfab4f08b012fd96ec8edba71e2ad020939e4 \
|
||||
--hash=sha256:1af817e84578ef8b7295ad17de0f9896e4c8520dbf2233c7aa5aa3d487256fc4 \
|
||||
--hash=sha256:1b0bcac4d490a237e18cf91f57352920c29f77f2fa39efd0813fb81298bf17ba \
|
||||
--hash=sha256:1d98e33bd8bd67d7206c124e200bf2229c4cfa8c9c19f7b44a897f0fc71837eb \
|
||||
--hash=sha256:21ca0396fe5ec995dd54431c32698189666f9224810acfa752e50d2bd94d9df2 \
|
||||
--hash=sha256:224865cbbcb7a2bd1356741dff12b0134df726b6d44bb7b500df8e303cbd9e81 \
|
||||
--hash=sha256:242bb0cda2ae3650764fc194593d9ea45fc9e72729acd89778c7cfe184cec2a5 \
|
||||
--hash=sha256:27f1821903e2ceadcb88ec2b45ef190897b7682449c772f4d9b53e42c520cf29 \
|
||||
--hash=sha256:28524438cd3e723f25412f63d4fd516ff5bae9ae5aa56acbe2a1404398a0cf31 \
|
||||
--hash=sha256:2b741888c93147444fdfc851abd81cc207f37f7f7da42062a00deb3888e57da8 \
|
||||
--hash=sha256:2c36ff87b5dfaa477d0bd51e9d7f6abdae7c8955d2983c97419085d842154b3e \
|
||||
--hash=sha256:34b7d9c24a4165a2c61cc8ae11d44d48c9ce2830fb536cb7914e11fdd9962728 \
|
||||
--hash=sha256:49d525938467d52c923a890153c99087c9d5a937d1f6b585dbdba34ec82e397a \
|
||||
--hash=sha256:4f84cdd868978d7b7350a566c254042d44216d9e37f241f3a6d3b1dfebeede35 \
|
||||
--hash=sha256:62ff20cd130c956c7c9144d5fe35228f98b51c579b2439e988b27ef93e16c02a \
|
||||
--hash=sha256:63505c71542a44b68b1e38060450fb006404170da375feb31af153e7f9c6205d \
|
||||
--hash=sha256:6376d4b3aca039375ca8bf92f770da0ec424a1ce3a37077a8d3c557411aa56ca \
|
||||
--hash=sha256:6a4e68eed961a8de6928d1c17ff3dc2a547e0e923c17f8f1cd79fb7bc9502f98 \
|
||||
--hash=sha256:6ab674f668d5962a3a4136ae0812519b0f1586874263723a32181d60d64137e1 \
|
||||
--hash=sha256:7014ab7e6f5d8511af92544667a0346ea6dfc314ea9a7cad1dba9fdb5c9a6e33 \
|
||||
--hash=sha256:76ae29acace5d33355344612844d588e19deaaba4639d8bb01601e4b1418ef36 \
|
||||
--hash=sha256:78de2d65e0b9ea7ce9d1b1c3e87297b2d7305a02c266ee2a2d6910daddd7ee69 \
|
||||
--hash=sha256:9bacedc04b0402837586a17f0919e3dfdd95291f441f1f56bd80ec274c2840a1 \
|
||||
--hash=sha256:a86c069c91a747a2c4e5c51473590aeb48172fff9b2130d23729a42d98665ecb \
|
||||
--hash=sha256:ac82fc756a446b6ccd7139ce70efa9d8bbe541e7ad579a12dcb52764b7175c5f \
|
||||
--hash=sha256:af11ac37a7c53dc16cb7950a6190851b0870fe218b6c60c0bb7ac355234e3083 \
|
||||
--hash=sha256:b70225b5fd1e0d2ef4f7fd30d24658454535f0924dff0caca5dc08efbbbadfbb \
|
||||
--hash=sha256:c49e853a3bef9dd10329f31f702e7fa9b5c58229ff9c2ff6d069efaf09177c08 \
|
||||
--hash=sha256:ccaf0a46cbb380f1fd102a874e32aa629fd3cb0c0e94f4943fa1f6d5edc5dac6 \
|
||||
--hash=sha256:d157ddfab1e8b21f2f1dedda9c09645d98b5ed0b667b0626be600a345d426440 \
|
||||
--hash=sha256:d88e5f7e60f28ae0b0cc6b2f16c43e87cd642a196a86f85e0d8bb6fe016fc16d \
|
||||
--hash=sha256:db0fcd827ca61622a01b220aadfbece01939acf53888f2cb98cd93e9b1e2c97e \
|
||||
--hash=sha256:df612391feca41c44d20118f3b88d1b86419465cd1f5496859f715ca60ec2210 \
|
||||
--hash=sha256:f0c3103fcff20183e593459cfea6e012281c0e76ae3ed8b5565ad1b92eac3990 \
|
||||
--hash=sha256:f9c4420a7a864fe1b86ce35befc95b8e39fb852493b81cf798671ddc265de638 \
|
||||
--hash=sha256:ffff613aaa9ce6236766e2fc6dc560bb5abde7a2e2416e3db1f9ae395a2b4dd4
|
||||
# via argon2-cffi
|
||||
cffi==2.1.1 \
|
||||
--hash=sha256:046bfc24911b37851ee1b51aab8bffe713d89c68c6a057b09484ce9fd5f69b4e \
|
||||
--hash=sha256:06c72bb76605a4b0cd0aad6930b69d4baf7dd5d806cfc409b824191099700e66 \
|
||||
--hash=sha256:0beceaabe56af686895136a2de78db54ecd8e4046b236b8fd6d6cb61389e9bf2 \
|
||||
--hash=sha256:154852545011f779917b11c78db2358d095da62a9a172b78ad0a583ee5adc0d0 \
|
||||
--hash=sha256:194cffa889098ced9976c3fc6340305e43f6303657d298da55366907c05c22d6 \
|
||||
--hash=sha256:19ee6127ee34de7d83ce3d371ebc5ed91addbdcc39f9ab15ce4eb35a4e534971 \
|
||||
--hash=sha256:1a18a57b58cfb21fc28d72e876acf10eaed67a1ed96226f92af4df681d571c4c \
|
||||
--hash=sha256:1aa5645c30469b09530c4ebca77ebf8f17618293c58f8549cb1a543a50236e7d \
|
||||
--hash=sha256:1dea0e4d7d4f11f619fe8c1d76caf49e24405b4b5743c0e3be16a500ecd930c9 \
|
||||
--hash=sha256:208f941bb9d18e768138677f0a6d2ce01f590df56043dda1df1535ac57c88517 \
|
||||
--hash=sha256:210019b6c7cf07f081b4c54635c8cf744377001350e29cc0f81c4377b4797735 \
|
||||
--hash=sha256:246fa40ce8645a614ff682e0b70f37134e460eaf93a775e0cbe3cca585a67a80 \
|
||||
--hash=sha256:25792eac27877609e7bb06d42ff88278a6624fff2ba9bbb523c09616b117e80f \
|
||||
--hash=sha256:27350daa11d4f10c540e6e89dada4c54feb7256ad03e9a4dc075ebad7ba360d1 \
|
||||
--hash=sha256:28907ab9bfb6aa13184cfc17c6b8e1023c5ab6fd7076d8c20a35e59fe04f8f29 \
|
||||
--hash=sha256:2ae64be792b8966f2c69538199728b290e34726562896df1e5dc8ffd8d8188e8 \
|
||||
--hash=sha256:31348097ff5bbe827ccc41795d4dd099d9f0625e7def00ee653c137a490c2a6c \
|
||||
--hash=sha256:3143d81e29e1e20a9ce10901ec369012947876596f75a222235965f2b7ae832e \
|
||||
--hash=sha256:3222ba5d678f80a030e6afbcc33dc1ae5cb45facabb61cee2c7016b8432fde48 \
|
||||
--hash=sha256:3311ed60d36f83378794e1009ac6258bafbf81f7888b4caa7b35a521e3f95813 \
|
||||
--hash=sha256:334644fbac4eff73d985a17a91226df55d0f394160c4cfb880e084c8f7161cac \
|
||||
--hash=sha256:34e261f78cb6ceaaa36f42f2613f4380d94d9c759a9c73c769ee6e0247364632 \
|
||||
--hash=sha256:363e05fa78e15116c3c32c210ee36884fd6b9afa6d440e47112c3bd511d64cb6 \
|
||||
--hash=sha256:398aff33cee2767e3e781d2554c54bd0dff386bb437581e0d8011fde1a942ec1 \
|
||||
--hash=sha256:3d22a20b1fb1632cc72c22f95f7b0d2961c3e1c235f245ba4c606c4771035659 \
|
||||
--hash=sha256:42a494cee34437f05546455144f2b5d9ac09b1face62bcfce597d2e521066688 \
|
||||
--hash=sha256:42e2f76b9455f5a9a844f770bf3e200ed3da0e15f5df3db9c31fe80b04b3d004 \
|
||||
--hash=sha256:42f6930c31dc7f50732c9ae793c2786c7b6b044195967bbdde40bb9be81c4cc0 \
|
||||
--hash=sha256:456a61fa52d579ebf9df2e9552ead5129855dbaff6c1e5a9b1bc408809bdc062 \
|
||||
--hash=sha256:471cee653ae88de62096552e6d24ccb4a5adb8c8c9f10b5054d0122c15bf2779 \
|
||||
--hash=sha256:49cbc70e6542d4ccccb936558d1064a8012541e78f821f955cff24e357776c94 \
|
||||
--hash=sha256:4a7c934f7360e8cd64fe9efadcbd10c7c6364f531e432b9a4bf5ccbc9e0e8b50 \
|
||||
--hash=sha256:4be96343e422f2dfcd12ab5c9f5aebe03f82f737c6bffeca6830b3875cb44aab \
|
||||
--hash=sha256:4f42141fc14250de6dde5ee7ea4432be017252d91f19c5ad043c084cea629cac \
|
||||
--hash=sha256:507a24c282e0f42f8ed737cf048572cbf580468da5555764a8331735e9c736b6 \
|
||||
--hash=sha256:51b31d1c98274844cfd7838ce00bfc27c7423a4dc00fc0772fc3331c2cc90676 \
|
||||
--hash=sha256:58acb8ab8e295e6c5ea12f888cbb13cf21511ef2a3303a23f4325c29d17fe5c1 \
|
||||
--hash=sha256:5a59cc1c4442bc3d5c703bf720b51138d0bfc173618807c9ee2490a7541dd3d9 \
|
||||
--hash=sha256:5bb4e7ea95dcd6a014a6fef62e62467d67d8e582326443f3d68e71d6320a9fcf \
|
||||
--hash=sha256:5c58fe613dc5e5336357eff555824a314d8e43282600435c8d1cb6a7a2fedd13 \
|
||||
--hash=sha256:5e7cecbaadb83884793e05828cee59b210b24583b9c7425d0ba6a754fe22eb4e \
|
||||
--hash=sha256:616f097f2fe415bc92a247f02e11f634e1f9e9a83d327e3c915c15089c87869e \
|
||||
--hash=sha256:63bbfd5ded17c4840ac07cd8f1c21ba9d9708141f840b324f422f41b207e3973 \
|
||||
--hash=sha256:64faea20f4e2613363a1a9b9c7dd73058f3ecd00133a511e72ad7c511658f527 \
|
||||
--hash=sha256:661c298b4821edebead0c91edd2b00374d67ad7c5a1f7a91d4442633b79d6a72 \
|
||||
--hash=sha256:68e62fe11f30d5ca8289242866f0a5291402d8529ca2178ab8afc5c9694ae890 \
|
||||
--hash=sha256:6a8dddef476fab96d066d578fc88526767b836ab5ab21754e1d5bf3879c31c7c \
|
||||
--hash=sha256:6e192623c49c94421616a5778fba35cf0d5a8d000650c1967ef4448ee5cdd990 \
|
||||
--hash=sha256:7225e4514edb64eb6740324353e0da0711954fd8d7da4576755b1c6e09b697cd \
|
||||
--hash=sha256:75f80557d1389eddbd0de2681f6a390a0c5338c31ddaa821381c203fc3fd50d9 \
|
||||
--hash=sha256:770de9db11e84213beec501cfcaa013b019820ca881e03344dea5844f7876d94 \
|
||||
--hash=sha256:7750c6449dff7864bb9bb27ddfb0267756189201a3afc911d82b3caacd70dfc3 \
|
||||
--hash=sha256:7bde5e4cc5c10140859842b9d383af292b22639a4dffb725314baf45968cef80 \
|
||||
--hash=sha256:7ce713ace7c0e4520535b42b77eaa742c16dab813978064913e5a3cf82973b41 \
|
||||
--hash=sha256:7da0c5eff80f0197f3b3d1232ec5a682a9325f4ae9016a78f5f5ca35f9ced1f5 \
|
||||
--hash=sha256:7dbb61fe3a7699468030f71bbe5f8a0e326a151daa91beb11a6fc1f980c55e1c \
|
||||
--hash=sha256:811bd1e21d32de12efca32393a0ab3f5133b54fce9bd44b8bd77ab07da14bf6a \
|
||||
--hash=sha256:8ef53b2de9bcb9197d31854256575d59dbac0cba72ac627bb291ef5eceb74be4 \
|
||||
--hash=sha256:937c0052c05a31ca1daf18de3158eed4dbfcb9cc107adbea227728d647be701e \
|
||||
--hash=sha256:9d2055050ea716bd38b7f7f1579c275386646b4894c155a3e2f3cd62ed41b7c6 \
|
||||
--hash=sha256:9f8d177621de5cb38ee3e731eda45d421db093ec0739f46a5594babda7987a98 \
|
||||
--hash=sha256:a2d7755bef5a12ed488f4ef1f1b69ee9191d7396083b755a5d2295f6edb4768b \
|
||||
--hash=sha256:a48d62ab9d6f4f98c983223a547af44be6ca3691074c31cecced6facd3ba2dc1 \
|
||||
--hash=sha256:a4f00aa42f75d6e4595e8866e748cc1705adc0cddfeb2ca86d0d03993d63ba03 \
|
||||
--hash=sha256:a6e721d4b0e45d5b65e87534470e67b18dcd092c83f68fba09f152b9cbc061af \
|
||||
--hash=sha256:a730a083190634c65cca36ba5f489531576ebd79bcd5c8e172130f6453127231 \
|
||||
--hash=sha256:a931079504ecc49efed7744c476a5c343a92fabf66dec2db95edb1b2fdc770e2 \
|
||||
--hash=sha256:aa9511c62d14da7aacc9b4bf51f3f697a621e83b2d6919008243c3aad168eea3 \
|
||||
--hash=sha256:ab36d55f9ed2d067327667c2fea18dda018eb628dd6347aa01dda6cf1f5d3836 \
|
||||
--hash=sha256:ad2c86c495b899d862ea0f4b42891b8713a3bd45dd4105c7fd51c2a72f39f3a5 \
|
||||
--hash=sha256:aeae0e330c9f6acd681f647d46cefd30c29f93e3392882e792e82080c9691399 \
|
||||
--hash=sha256:b0431303acaea1089ad4b3e9ce4e6518193def1118d4073ca848635ee4ea2e96 \
|
||||
--hash=sha256:b5bdfd1c873d4e093aabc0ca84c4ca6dbc4f752afb5c86f146d9742580c9da2e \
|
||||
--hash=sha256:baed1e86cc735622097354b9d1281406caf42ff42a886d29faa8e8d1630333be \
|
||||
--hash=sha256:c1453022f490d2459a11819d83ad1d586e9ff65a12ac3e705ffebd46d3685dcf \
|
||||
--hash=sha256:c26608d2222fb1e94487e4a387d85f13eb55d5ed725cb25a0c589ac4ee60e7bc \
|
||||
--hash=sha256:c7659f22557c5a0bc4855cd635f55edec690cc008a40768527762cb9fb263455 \
|
||||
--hash=sha256:c8c69575568085ba0b1b10c0249d779a214aea6f6522e949a0fc9fb0fcb449d0 \
|
||||
--hash=sha256:c8d2c9fd1f2d16f780d15127abb050d13d1a76c03a4bd87d7e4980e45e511e12 \
|
||||
--hash=sha256:ca82be1a1d406ecfe1d25dc16cb33488e5a16bf4438c9fb590484ea29d92478b \
|
||||
--hash=sha256:cc572dace3f60ef98d7b12ff411d20f5362feb31a0439eab0085bbfd349982d7 \
|
||||
--hash=sha256:d18e5ac0f2f03f4f518d3e23db0f0cad7faa1da8620e9c09461d443bbf6e6692 \
|
||||
--hash=sha256:d28630f5854ab07ab1fd4aba756de52326c82e6be15d414b12793f1975048b54 \
|
||||
--hash=sha256:d9c275eaacd24aa73f94ffd6de08fc3f932424d8b6c376f4bed7cde376fe7bc3 \
|
||||
--hash=sha256:da0e573f9f97159390c89d9f1a9e41908b66d408cc5b58d08cf3847d844c531b \
|
||||
--hash=sha256:dd31f52ea1086513bb9df30f8fcee9b8918323ae067a3d5b78bc826a000712be \
|
||||
--hash=sha256:dddad92b554513a31f272570678ba307fb9f618f05e3d4a5eacafff9eae03e1d \
|
||||
--hash=sha256:df423d40ee8654634421812bc3b196da3f9bd7d32929da813f8394c4348a5358 \
|
||||
--hash=sha256:df913725b79db7bcf03448f36b7bf8815363417d5b58deecf9305e3e30f0f21a \
|
||||
--hash=sha256:e0bcb7e0f677f543555d2adff3bf19c05f66cdb4796e5ff602442ab2fe3c4ef7 \
|
||||
--hash=sha256:e2d65b31f36619cda3999b78b2aa9632e76b78448e7a56fc4240824200e7c4fc \
|
||||
--hash=sha256:e6e8cff14d6fb0be70a09c0bdc58096f501952d04624ebf867e0e56da2df8960 \
|
||||
--hash=sha256:f16c709686a78c727bbbf059f92b0bf41c6fc60deec706d2dc19f529175a6125 \
|
||||
--hash=sha256:f24fb43132a4c6b4cb4eb029492919b2db645be6808d738f244fd146c03c32cb \
|
||||
--hash=sha256:f53e442b08449d42821fa4a4fba000095af9f62742a500f978a9f557ec44339a \
|
||||
--hash=sha256:f5cfbc5fe74540d335175b656c725d74d90e3730c626d92575eea35029d9afaa \
|
||||
--hash=sha256:f81b3b8f3d4e343550fa4baa0e479bba9f2d29ce9c2e9b51d1ce1718d7442fcf \
|
||||
--hash=sha256:f8ec5e643a9a937f64e1999eb9f75d072263751912dc5cd06d3c85f8f44be7c3 \
|
||||
--hash=sha256:fb92203a88b3d3053034db775110081c49d28be6551923805e039924093761e4 \
|
||||
--hash=sha256:fcd22650c908d7b7da162bbfaab594a1227a15d1643a98c68b122ac642fa2264
|
||||
# via argon2-cffi-bindings
|
||||
click==8.5.0 \
|
||||
--hash=sha256:255bc9599cf7748b4b1a446ccc735421bd08a2ae529a8b88597d3de5664ee360 \
|
||||
--hash=sha256:ba0d2089de75ea0310e2dde03160e6ca10009947fb95a182f9b54021bb272e34
|
||||
# via uvicorn
|
||||
h11==0.16.0 \
|
||||
--hash=sha256:4e35b956cf45792e4caa5885e69fba00bdbc6ffafbfa020300e549b208ee5ff1 \
|
||||
--hash=sha256:63cf8bbe7522de3bf65932fda1d9c2772064ffb3dae62d55932da54b31cb6c86
|
||||
# via uvicorn
|
||||
httptools==0.9.0 \
|
||||
--hash=sha256:02bc5b3dcb6394b9d825fd62a7bfa0b2943063a3c89abc4492ad45e334a20eb5 \
|
||||
--hash=sha256:050f7ab098121873c8f13e35857f97ab60a76185c8302bde9a384939bb7c3b96 \
|
||||
--hash=sha256:050f84b7ec46a6efe0e5f521cf8729e3397c1cef4384f62ed8d5d68ca0045776 \
|
||||
--hash=sha256:06bfe7fad972a417269d8a5fc53b87e4eca970354abf5e9e24336fd06d64292e \
|
||||
--hash=sha256:088de1738e1af624466a01c35d652dbe6fb825be887c76d68aa850621d81db88 \
|
||||
--hash=sha256:0adc974916efe1fbf89d0363a86dcb2c746727643e362ff398de1a4b50b6bc77 \
|
||||
--hash=sha256:0cc339a807c156d840b54f8bf050ba0fc265eb81692c24bca8535b52fbd797c6 \
|
||||
--hash=sha256:0fd73d0bbf700a30dd87e4412adf41cfa71542a533d6b390c7244bbb8a1152bb \
|
||||
--hash=sha256:130635fea6e611a6b2026120037965ddb88b3dafd11bb64e264b101a70a76630 \
|
||||
--hash=sha256:13873eb8aef5972fcfee614f63d47064312ad4efbfe65ade15b8a3b77f8c8659 \
|
||||
--hash=sha256:18d800aaa2d6bff7d889df810d1b19a5fde72b1f6c0ca96e8d9f28a692fe5460 \
|
||||
--hash=sha256:1a4050a651e1f2faf05eb028ce9f2168abbcee9e24b209f5c1f2eb96d8c569e4 \
|
||||
--hash=sha256:1a7f1df31829c258158be01bb04eb668c4fba7df1ddf2262131a972962e651b6 \
|
||||
--hash=sha256:1b01c0fcd6725a8d79a164ecdc4116866282479d68bb3d6d74a909bf994656c4 \
|
||||
--hash=sha256:1b95775f6292d72cb452c33e5c0f8b8551807c29a10e3c1671fef7f61361370a \
|
||||
--hash=sha256:1f6da814aeecbc6cb8872d6d3e85ed16e8ab1653f9557cea8658725ce212348a \
|
||||
--hash=sha256:2095207b75a83c9e947346da9c127fb7e4fb29f41589df2643764f06b750989c \
|
||||
--hash=sha256:22ab1b10b06d357f01092e60f5e6856a0d479ed79b0ec2166a339ea26c699be2 \
|
||||
--hash=sha256:2319858018eedd0c0b2f950a620413c0a9d1352607be4267eb28209eca8b1e3f \
|
||||
--hash=sha256:268d18601feb5367885c6ebf6f402c18fc25a324cee215784adafe0a1eef925f \
|
||||
--hash=sha256:26e1d9629f3bf70d23f0d22238152aec51c837a7c9e384cb74f356fdccad7eb3 \
|
||||
--hash=sha256:272db0c51e8b71e953c1f2ecbe63402b819680e4564be2ef285cfd4584ee8355 \
|
||||
--hash=sha256:289f213d2a3dde2e8312c415ffecec5a01698589ec6249ec4e8fb3b47c0444ba \
|
||||
--hash=sha256:29b0d823e3c1e7cd1093a5dc889245db693ef13ada624cd66e2262421ef38867 \
|
||||
--hash=sha256:310266a2db1377ffae3bdf6556ab4973f4f94508a8ce37b2f6bb096a89bcefa1 \
|
||||
--hash=sha256:3238e198429cb8909ec42951b82d6a33fe0fdfcf86371732f8f09311c5b8ac32 \
|
||||
--hash=sha256:34266cec8c1d4e3e91fcca7efe38971d6bdda64a7944f2a46ab576da15173680 \
|
||||
--hash=sha256:36fac804b8cfd6b935ae64f71349f833d2b6298404626d017a2c57bb942bc643 \
|
||||
--hash=sha256:3af4e45ff455fce5511fdf2653c1ce428ef09c56fe37a83eb4d924c2d474f31e \
|
||||
--hash=sha256:3e3201fe4d46e0d15d7ff9fafc94a605da9eb82d2c5b9837f0368acb325481f1 \
|
||||
--hash=sha256:45b3002392948dcf578029c89f6318e1289a993a1a5ec38a4161560fab60f811 \
|
||||
--hash=sha256:465bc1526debf53a3be92022a16ca0c38f891ea3b5c1587af4f52e44020f8a07 \
|
||||
--hash=sha256:48c705bd0b1afb6253ed71eca9f9ba7ac7d47838e5fed1ef7891d67f21ecd4de \
|
||||
--hash=sha256:4a4d8c2c7e73ba5967be74d7c3a5ff81fde815ee1b48d9c5c0f14de8463a847b \
|
||||
--hash=sha256:4a85401b0c3f893cf5695c1199e8679fbf673f7f78c2f6c11d6b1850f8c7e358 \
|
||||
--hash=sha256:4c58dc91aefb31adad500aa68054334f429b840b36dd29e34e834101044cb2ef \
|
||||
--hash=sha256:4efbee349138a3fee7a4cc3a95abd2d499fae70dd5bff9fed9138d6f570f4283 \
|
||||
--hash=sha256:4fb995082fe41ec410b33c48b54fb1d44abb8a6ee762c31e8c42519e8c3a30a9 \
|
||||
--hash=sha256:5042aa1c7e2b1a24c17dab31d8770b63a5101c9abc25f832c6aef6b201e1ca4f \
|
||||
--hash=sha256:52fe0176682a25b15370f23f5b0f1366a84771df89144fb0cd979cb72a94b5ca \
|
||||
--hash=sha256:5332a020a60bbe32ede4bda1a62b3d56c4831d309cdf0932842c0fca8ad6aaa3 \
|
||||
--hash=sha256:563e4568217dc907a91843f38c737be865222c0400a38cdcd0d26ce92b3db271 \
|
||||
--hash=sha256:581b27663c6e9f4df68068f32fe6d1cd7647b31fac90237221a66f8821c342eb \
|
||||
--hash=sha256:58a1b0ec4cbb930e69669f9771715b2c7898d3cdf064d9811f7a66afef96b544 \
|
||||
--hash=sha256:5cc5d3a29f9ec86ce406e5ec09c241dd8dc4d30e838f74f68d728b89131a3acf \
|
||||
--hash=sha256:63d38e9a9a10a20fb57593742e63c6b1e78dd7f6ef5472de8e0b1e4cf4f3db26 \
|
||||
--hash=sha256:6b1ac7f1bc6c0dbf90684b77571a51a21b2463909fd916ce0ac9bfc4d566dc75 \
|
||||
--hash=sha256:6b900073e7b8481ef1aaf4f6c1789d210a1db01a9da8789821578cfeb4c2d540 \
|
||||
--hash=sha256:6c12d0393a903b58bc5f5a7406d6c5290acfb8284290d68547ce620c06f7d133 \
|
||||
--hash=sha256:6e2780e33a58a93f27cc3bb74a55bae6f9a8278a1dbabdff392940d30d381671 \
|
||||
--hash=sha256:6ebd39ee26db460cfe5ab8b71a15d1149b289139a0d3981522757d6af620887e \
|
||||
--hash=sha256:6f8b41299b203ce8f627db670cfea82067d9638853dbeaf86dccd93878879b85 \
|
||||
--hash=sha256:6f9549ca354a1d6d6167c458a1f1b12147726b968f02dd64b6a5801dba91ae0f \
|
||||
--hash=sha256:6ff0145b34610e57c9fae20df4e133c8d54266447387de6fcc0bdabfe4db4569 \
|
||||
--hash=sha256:6ff5f0ed70783dcb9562dbd20edca51c3d4d277f128223709e3da6b75986d1d4 \
|
||||
--hash=sha256:714bf348f468532d86bed670837e7d5ddff3834dd7f5d3c08066da400c86f088 \
|
||||
--hash=sha256:757e3f79cb865a7db94e0db5f4d0ed3284a69e39d53568f433982ea13c60cac1 \
|
||||
--hash=sha256:7e32b83bd8c2f8b6fa726ef34e63e21c4d7eddc277d40d4ef7245ea3ed28e5b6 \
|
||||
--hash=sha256:805b0f2618e5d4c3e28f45b731eb1a0539691ae4a2f97b4ce014de0bf96a1ff5 \
|
||||
--hash=sha256:80eae881cfb69383303e9a4d7961a478025b89c24f38f2e69b30c516fa0d57f2 \
|
||||
--hash=sha256:813a32f94991b9627795528053c73a57d2ce3eb98ede89f0e1c7a31095938e81 \
|
||||
--hash=sha256:8463b34ebde3f000627e9dbd8a545f995ad49fbf7ff9dd5abc0cd507da98a603 \
|
||||
--hash=sha256:8a59c749a73fbdbc8e63b895a3079825fa085d752e75bc0a500042cb8a801e48 \
|
||||
--hash=sha256:8d90d10e9b6594c28f27896a68fab97fd784c43804e9fe419dab8e8dcfcf4b02 \
|
||||
--hash=sha256:8e1e037bb57dbc549c6fe20370b763ea74bdb09413cdcf857e4f14d9e4e2fb13 \
|
||||
--hash=sha256:931f45f84e15daafec5f82cc92e6710569e1f50933f3253d206eab4132bec678 \
|
||||
--hash=sha256:995b52f7c260ac7023640221f27472303968753cb6fc6fce1ddfb0e9db59a398 \
|
||||
--hash=sha256:9b4da5789d7cf576c7e81f0088c632f6ee3786d87d17f08e90e703c22ce15633 \
|
||||
--hash=sha256:a3ed60ea9a7c352c590182c67404599e6b5a0c901e75ae4cceee9a9fd6bfa455 \
|
||||
--hash=sha256:a4d1ecad62e83cc65b411ea0125972cf3af98821e8117129947fd1e3a113f8d2 \
|
||||
--hash=sha256:ae9bb62a7902e2ab65782447cd3eeb753510feace4e3ea03937a85489b01b16b \
|
||||
--hash=sha256:b2ab3aad55d75d0b8df8d8a1b5920baaec9b161112cd5e95984848b4d2cd3dfe \
|
||||
--hash=sha256:b2cc6991f16f6d666d48e4b57318104e7b29109e32e2f6b86e9d44c4e6a27f4e \
|
||||
--hash=sha256:b5a3f5f70967a1aa2bc47fec42a1e19d2fb38c61700e3ee62b63a4af4f4fd001 \
|
||||
--hash=sha256:b68fb053b37c258a473ab67f4965c3b439500dc160fe364667035a6833eaf50a \
|
||||
--hash=sha256:b6ee42112d785a913dd63ec0335435a3dddbea5040c151252db815b0095cf066 \
|
||||
--hash=sha256:b928ab0ecaa664e8caecc529dcb8bc881b6b35bb2b74bf9a39ae25f982ee8812 \
|
||||
--hash=sha256:b9430f65db521db7962ad951571d446171213686f96c998a54dc18ed574821e2 \
|
||||
--hash=sha256:b9cd15cb7cf0d5cc41f649fd789aae12c56c3b83eff593f8e095c1d4555ad5c3 \
|
||||
--hash=sha256:bb1533541c729ad422f870a780d8b4af924f9817d45b5f580390418cda72eaa2 \
|
||||
--hash=sha256:bbf7377fbd41b7c87d47820e25b9876724963681c2a1d6f6ff2adb4db46ac174 \
|
||||
--hash=sha256:bca180cbe84e4fba7807eb408a8655295f697928512324517e30a091ede522a8 \
|
||||
--hash=sha256:beb2c8a34cc90fb4d862b7284eafdb322030d6a8b2ee5eb6a744f84205beedc3 \
|
||||
--hash=sha256:bfdabac0c6d3d6a5be8c2a100a001c92c14a39bbafd5999545a675c493626e64 \
|
||||
--hash=sha256:c0e45def4d9ce7073e2226535572442d9d6efb4047c7a5fd8960807e877ce70a \
|
||||
--hash=sha256:c0f537e5e8152e8d9cae82804024790cb973061abd3b7ef8f66f46e2b5c7bb51 \
|
||||
--hash=sha256:c195a69df0ab2541252ab5b1d76e3c182e5688ac2a9b708e5e6f66aaeda91e9a \
|
||||
--hash=sha256:c271bfb832be5c5c020b4e2fcbc1e70a0b990adba6de874b0bba1184b89cdea3 \
|
||||
--hash=sha256:c42424213c28804f8d0e20f5692106cfb57bf72e1dbc4092b8481fb2f9e4c707 \
|
||||
--hash=sha256:c4fa57d3c31889722f64bfa785545a5e603a893b6f29ac1a41bfa830abeaefd5 \
|
||||
--hash=sha256:cb2bb3ac0af7fdab2311b895c9eb95442b45deb14cc949b9e65545e74aa0be69 \
|
||||
--hash=sha256:cb3e7a4fd0168e362673a980380bf4fd6ae3b1555150e60c5390b4b10d9c50c4 \
|
||||
--hash=sha256:cbbfcd5d15056fbd1edd5e725cf3feeb47c7cbccbe205927ebab422cc229f417 \
|
||||
--hash=sha256:cd3e55223a77d6e08d5730ebacb4930ecca5d2ce7c57e7ba10833be7e52903f1 \
|
||||
--hash=sha256:ce8e723b4637034b76f5382a30a6b725518c332273e8d62a6c7d46e90837c947 \
|
||||
--hash=sha256:d1e329a1866981efe0201d05a374617f6c6cf14434a501d78ab22793d1ab1fa6 \
|
||||
--hash=sha256:d20ba5c84cf0592afb2713336f07e2b6ced082e4ae803ceada153a85613efc9f \
|
||||
--hash=sha256:d2b095129b9a98eb46a271ee9631089529c4e40354576b4aa74e24de9d2bf2f7 \
|
||||
--hash=sha256:d3906b5c549ff2ad2473cb711e1fc65d76715c2726a402108fbf55eab6c6b49d \
|
||||
--hash=sha256:d484ebb7e3a3f3597b0f645fbd1b85633674ca808c1f5ba11c2caf7c66f5c8b6 \
|
||||
--hash=sha256:db735a23ecb0f0450d2b24e0a05fb00a8a35c9db172919c4d3e023e7c7ee4c9b \
|
||||
--hash=sha256:dbc9fd1521e573045d71b6afab7398439c5cc259e8cb9d416fe62d485c4899c6 \
|
||||
--hash=sha256:df3867518b205be3648e2fbd522bf380c851b5c2500588047505afdd786b6669 \
|
||||
--hash=sha256:e0acbd474d0af4afacc6e66c4273f8a19e25f8af4379fc816388095ea6b01371 \
|
||||
--hash=sha256:eacf0f45ca3ff84c01481c60c15da9ee56711f7292f66663df0f57af61e011c2 \
|
||||
--hash=sha256:ead1a40543a033a6732a9e1e515944979a19db3737ce77363fc0660e38554344 \
|
||||
--hash=sha256:eae4e9c7a0785a1a715de0a74fb822ab40084c060f444f18f075d05e322aa7ef \
|
||||
--hash=sha256:ecf7037e491c220cd73987838c1ac3958d787bb098c3be0bfaf7f04204a6162c \
|
||||
--hash=sha256:ecfeee649184ffd800955068be9a6b579a0f33fc3c98535d685d5779cb59347f \
|
||||
--hash=sha256:edd5aa045fa3cc57143db018dd32ce7962bd5b525d05230709015d7e570100aa \
|
||||
--hash=sha256:f0ef48ce353f6b6a52232ba23d0983d4c2c84c84a778899404e34b4718509bf2 \
|
||||
--hash=sha256:f1734bd6f588975ffc246211e8b96c11933344087ca280d2cbcbf35cf835d7a9 \
|
||||
--hash=sha256:f67db0ba2bedafec15b8e5330d40da1e1c7921559fa715af021252bfef81a6f8 \
|
||||
--hash=sha256:f6ac1414556b910a879c108d79736f77e797871f9919ed0d2c3cf8cf3ecca986 \
|
||||
--hash=sha256:f78f7ae1c2e5aabf29583fc0d302d8081a663776f84578025662eb6f5d63a921 \
|
||||
--hash=sha256:f9489c1d87160c126f73b004742fe8654fa1ce37ed89e9e01330a1c10aaecde4 \
|
||||
--hash=sha256:f9ccc9884241efceb4547a92955d128574c864681f11b7ea3ecbde295fafbe8b \
|
||||
--hash=sha256:fc1a4f9d18d32a6e0a0a0a382986a60a2126f5144dd08715be7adb8df18e8a46
|
||||
# via -r requirements.in
|
||||
idna==3.20 \
|
||||
--hash=sha256:a7db850025b95ded1eae8a46181a1a6c56c92c96f0e2b005d9ff8dc0210cab44 \
|
||||
--hash=sha256:ab7ae7122974553370f0bdb919e1a960b2cd1bc1ef0276416d896db81c14582c
|
||||
# via anyio
|
||||
psycopg==3.3.6 \
|
||||
--hash=sha256:a1db9f7148b06a28606767efaca51fa6f9398c5c0a3810519be69d7000bdb631 \
|
||||
--hash=sha256:c081f2250df751a943036e42db6df4571c66cd0aabe8291a7a506512b12007d2
|
||||
# via -r requirements.in
|
||||
psycopg-binary==3.3.6 \
|
||||
--hash=sha256:05a83ac9fd52b9bca7cb5ab04b3691163170bd16f53defa27216ea3aa07ee781 \
|
||||
--hash=sha256:0a52991594ac4db888c7d39bccef331797e30cb31a95cae02cf2607f83a42dc2 \
|
||||
--hash=sha256:0bf08b749cc144f33b44a91b78e3f71c60eb07963746a0df5a100b36ce3d7475 \
|
||||
--hash=sha256:0ebfad5d131de9f892ae9e70cc7616207768b6714b66a52d4612b8ceaf78b372 \
|
||||
--hash=sha256:1679a1cb93fbe5a6d1fd58d82cbddcc6fcb8c61446ba7cae6eb2a7b19bc585de \
|
||||
--hash=sha256:198a48e68cc99ccac03ba95ac857e73aa66f3bf6be77019fafb0832a05f7ad03 \
|
||||
--hash=sha256:1fbd30e537dab22cafdf080608f10148fe2a5f3a61294ddb5113caac8a623840 \
|
||||
--hash=sha256:289aadd6a00e151203c081f708348ec89f1e483c9b510ef4ac3981f847f01f79 \
|
||||
--hash=sha256:2f122603f36050937982abf9668d8bc4769a79f7c93a65013b1c49f1cab7b56b \
|
||||
--hash=sha256:303732e798fe6729f8e12021b9c96107df8e95ecec4dd487c67b98ec2a59435e \
|
||||
--hash=sha256:31cd942c23f613276b81a6e6598cefa12960058b0f46e1e874b540c793f6aca5 \
|
||||
--hash=sha256:366db6e97e66b37211475f20c4c1324a2dc0dd825e46d4e87f9d599304d276f9 \
|
||||
--hash=sha256:373704aea331d3f3e3402c125a1543f5875e2986ebb54f97d1647942161f803f \
|
||||
--hash=sha256:37d40450659401600e6d043ff586c89a71a69f33cbb8bcdba6cdb2569beecdbe \
|
||||
--hash=sha256:37e517c146b185f9c0c6e8d0a0ebbdeeeb67896af28466e032bc810d0c7dc7a7 \
|
||||
--hash=sha256:3af90f92769d8cc10f94515ee7a0aef36ea85ca733a0ce22858f6e0953f41138 \
|
||||
--hash=sha256:3c9e663b2e800e3218994cf948c11bcc2844e6491b34aa80d089baf6531827bf \
|
||||
--hash=sha256:3f84dab25e0385692ee13274c68678377e0b1a70ab9d14e56264cbf61f60c62d \
|
||||
--hash=sha256:4690cf67738f0e0e49a32aeec99bf0e4595cc2b4f1af984a4345394b1dcff91a \
|
||||
--hash=sha256:566dd827f17728efdf7d88a5b066f815170f6fdad13967ae952842d90e6aaa9f \
|
||||
--hash=sha256:5927b7ba63153cd8e9862987290a2b783a5c590daf2a4ef981700cc3569166d4 \
|
||||
--hash=sha256:5ad8f35e67cc16d1fad1fa8c88972dc9b3a3141ea67897399904edab96a301b6 \
|
||||
--hash=sha256:5ea8beeb5541780b4b50b462eeacbc4f594ce3b911dc20c81c75f267876f71d2 \
|
||||
--hash=sha256:5f598f19fa9a91540b5cee17932ffd227b7b53a481605bcc4573c0eafa647300 \
|
||||
--hash=sha256:612382ac3ed13651c7fa44b5fee9fbf7baaa2ddbc6f500391672682c5f1df9e0 \
|
||||
--hash=sha256:6ff05561e4a067d35507dc5c90f1deb2ec1c9703ac5cccc1bc26e08a197f9c5a \
|
||||
--hash=sha256:7308c93cf0b19bbaf8e6ff0a6ad50d3c442385739245fe15a8d593bf841734a6 \
|
||||
--hash=sha256:79a2a1c3449f6c3409427078ed1cec10de79f3023cb5f2504f0597d350ad46c7 \
|
||||
--hash=sha256:7beb3e41c9a1e509f3ed85263386588cbe3e975aa67be21f79f44fd35ffaeefc \
|
||||
--hash=sha256:86147cb5d140341c3363fb5bacce31f8d5543902a46699d3c536b101bbceaf9e \
|
||||
--hash=sha256:889e42acec10450185e0cdfb396f375e2c1a8d7737c114830a7fde4654f59e30 \
|
||||
--hash=sha256:910ace140e3e7b7596898d083f37a8fe90c5c40684252ad4e682364b2cd3deba \
|
||||
--hash=sha256:955e3dd94da361e052d2e49acf591017158dc8f8ed2c8a42c2e3943403c39dc2 \
|
||||
--hash=sha256:9892188bb15e5803beb51afe8a25add6b56be391a53058e8bca03b74e1e6bf22 \
|
||||
--hash=sha256:98c02090d88f2ebc0ec1e8da538f77d225ce0fffecf372aa39262e62a1b054ef \
|
||||
--hash=sha256:9b2f11794e017ce340934e35de46181c46ef71ec75ea3d85dd75cd836761c01e \
|
||||
--hash=sha256:a2e44a342d2aee40508e28a563d8961c39d9bbd8cae36d8578f0a3c6658aab0f \
|
||||
--hash=sha256:a4ee3bdd5468a725f2a4d9aab8a74b6d0279f768c8b5d3aeb102c5307ff3d59c \
|
||||
--hash=sha256:a5165300324efd5a772c48a88ab3a928513ab3979fca76553e62ee815f7b2b9c \
|
||||
--hash=sha256:a9348c5b43a3bb5ef8c2e89d5237c9c87eeafb01d338c84a7aebbc5cd0313299 \
|
||||
--hash=sha256:aa73160077345ec21b3f51e8e24b3de2e99586217e497629326eb9b2ea88c52e \
|
||||
--hash=sha256:ad1c785e784cfd87e8436c6b7702f2d321fc39601bbaf29bc63a41a867091638 \
|
||||
--hash=sha256:b3f75dee0f9afafabe4edc52c4842f1e1878ed2069bd05b22d6fe961e97e4dba \
|
||||
--hash=sha256:b599defe9190b17e9907c8b4d114c181e702c87efcd1b8a0ad40971cdcc4634a \
|
||||
--hash=sha256:b82491019b884d62318b5f30706c3d7e6d4e5a6cb7eabcb3edc0c1b0fdaceae9 \
|
||||
--hash=sha256:b8ece331509f7a975b90501f41e83ad905e4141753fedf3f2711b2bc70a8efbc \
|
||||
--hash=sha256:b979a42815410432420275412633960807178b1ce26591a16ce06e78a5bd4bb2 \
|
||||
--hash=sha256:be4f9b3c9338ac5dd217c5847e21521b396c8117f78dc420d495a5c49bbef874 \
|
||||
--hash=sha256:bf8c8481d026b85dd70c5fa7dde85b2333aed0b32a2602bcd38a900cbd78a49c \
|
||||
--hash=sha256:c61617eaae0112ca154da87ffb99b73af2c74067acac28dfb9a4455b019dff2e \
|
||||
--hash=sha256:c6d19cb4999d03231e8730a5f66c8f5068bc3b532677eb39dab0f600bff3e312 \
|
||||
--hash=sha256:c7753871eb57e6a5f4646f6168590c6653073dea5e9e720b201c8875332df4c8 \
|
||||
--hash=sha256:c7f92daa0d2a1c76f07264abddf8cbabd30152a2f09c3270e50f0c7efdf5dcac \
|
||||
--hash=sha256:cbd5f73073ed19c378d4c35499db1e3e703a5b1a324e521204065967bfaa7a18 \
|
||||
--hash=sha256:cec5ea900390897d0b46130f60bc2883bf19c314f9044235217c8be88b0ef269 \
|
||||
--hash=sha256:d636338c8f21b0df2f84657b00bc34f9313f826ef93f1155bc743607e4a0c5eb \
|
||||
--hash=sha256:dc75da5a20951049f7b773145f998f69d181adad9c58a0ff36e0cf1d73c10e10 \
|
||||
--hash=sha256:e23a66a763fbe83fcc210bc77c27e5a5ea380ebf091c06f34d8561b695e5a40f \
|
||||
--hash=sha256:e8cbb54454dbf1bbf2ff08dd7693e8d94ac94b1a20f70f4b3b813d52ecb5cbc1 \
|
||||
--hash=sha256:ee2c4728c691245e24501fcd7a97b5b381236b9985bc445bba88cdce7d1b5784 \
|
||||
--hash=sha256:f0535693ce476a722b718b002d5d2c27d47e71ca945276ac194409c98e74c492 \
|
||||
--hash=sha256:f19cc87343eaa55255e76b31259a570072ac95d6ae82c92dd34b97691f5e49dc \
|
||||
--hash=sha256:f21d057f3e5f5491067e5b292498073b73847d48799b099803fef100775fcc52 \
|
||||
--hash=sha256:f87dbdc42e78ee0f7ea180c03f8c78e80a949e373066629bd90fefff10552dff \
|
||||
--hash=sha256:fa34eb47969297471db7b7f193622c7e3ee839ec05abd05f1fe104d5b1b1dcf4 \
|
||||
--hash=sha256:fdccb3a0e184b03e9baa673b15a809cf36c339c85dbda0ebc25a698846dfbee8
|
||||
# via -r requirements.in
|
||||
psycopg-pool==3.3.3 \
|
||||
--hash=sha256:9b9cd6a4fcec47a410f7e82d408540e7f77b478509e91b44c1a5457a13e5ff37 \
|
||||
--hash=sha256:df87b5d9d0ad7db37f6cdad4fa8ce113d250f5997f6db38e9a99192fb67f9e1d
|
||||
# via -r requirements.in
|
||||
pycparser==3.11 \
|
||||
--hash=sha256:51d5a8ba2be0bbe440b99d2112604c95bbbc3c2748a64260186c541e1729cd80 \
|
||||
--hash=sha256:d875f09c3507d00e1aba0eecc6dcadc1352f30fff09dc6bff2f1c2935e97c2bc
|
||||
# via cffi
|
||||
redis==8.1.0 \
|
||||
--hash=sha256:6e1a19beef9225c83efd689c7e6b7da2d5215b1f42cd13b7fc3714d0a09c7b25 \
|
||||
--hash=sha256:a4fe1aac3d3b3cc791d4b3d5931c5a956045dc951ee74d1c913ee3ac4d2ee9fb
|
||||
# via -r requirements.in
|
||||
starlette==1.7.0 \
|
||||
--hash=sha256:67f8e99895493dd2911a03f11314af6ceebeae4e704bb9f43dfc6a9db151c93e \
|
||||
--hash=sha256:c79f74ea63cff761804fbbfb182f1e0b440c2d07b164d24700c5a1bab5d6ff5d
|
||||
# via -r requirements.in
|
||||
typing-extensions==4.16.0 \
|
||||
--hash=sha256:481caa481374e813c1b176ada14e97f1f67a4539ce9cfeb3f350d78d6370c2e8 \
|
||||
--hash=sha256:dc983d19a509c94dba722ee6abd33940f7c05a89e243c47e907eb4db6f1a43e5
|
||||
# via
|
||||
# anyio
|
||||
# psycopg-pool
|
||||
uvicorn==0.54.0 \
|
||||
--hash=sha256:505bdb0f318731d45f1f712071fc781a8981f6847a31c902c9f5e652d4f67faf \
|
||||
--hash=sha256:a2e33cbfaa0306f8e6b0c13e0cb89d7d7a2da3e62b90c66e18c33d9807b28620
|
||||
# via -r requirements.in
|
||||
uvloop==0.23.0 ; sys_platform != "win32" \
|
||||
--hash=sha256:0305871ac712f54b62af73f943dbf21ae3ce80a44bc0f0151424484affa85645 \
|
||||
--hash=sha256:090865d8ce7a03986755a3ce711b7dd0d4b44eb14ab74368b717f3fad1180208 \
|
||||
--hash=sha256:098a85e1393ef5202767b7e5fb41a32cd8bd81e6ee4af364c179801c4aa3f6d4 \
|
||||
--hash=sha256:0efdd55bddbd36bb2fcb842d64c0d5f6407c6958c68088cc25df8c09edc5b5fd \
|
||||
--hash=sha256:12634f15e6625f78b3f2922f91404c4d7173487eba11746764153f556e9852dc \
|
||||
--hash=sha256:1748321e3c59a14a75404b1ae8d5a8d81c4e201803ea0e14c1b6fd84421024b5 \
|
||||
--hash=sha256:19c64108b507cd0bc140e400e3396bacebd9d504956aa7726272bf6de7d9aabb \
|
||||
--hash=sha256:1e84575f11873c109cf3962ad0bdf679094466184125f4cadcc41a73febff41f \
|
||||
--hash=sha256:24c58ae4a83e93a04c504bcc678125e36a0bfc44af928ad69444880c60f187a5 \
|
||||
--hash=sha256:28d160f51ab4da3b187063652e643dea6831072add4adc1e6d62afbe73b6be27 \
|
||||
--hash=sha256:2dcff2d69be43e6559e5dad2c5a7a2dbfb60e05a77311b6c4b7a4a8123d86c65 \
|
||||
--hash=sha256:31e0cf90bc8fd88784f6802cdba968a51fb1aec1cc3feec74d862b2d371d1330 \
|
||||
--hash=sha256:378188efbb1524f2219d05246a3e1e5907217848d2882144dff59585f1b81d55 \
|
||||
--hash=sha256:42feced24b9b44b856c633eafb5cc5dec354972da55ce77598db6844c054bc7c \
|
||||
--hash=sha256:4448e9124537620f9c25d004c227bb5104440b58955c19bbd312d910af919a63 \
|
||||
--hash=sha256:4a08875543bbd4519faf30497506c9cda8a48470467ffdf967c7313c7a5981a8 \
|
||||
--hash=sha256:4b8e207c67d207a8608fec57e116511030af3495dc0109b8c333cf9cb412b16f \
|
||||
--hash=sha256:4bb7f5d0b62b5afaaaea2b7b60d508921c24b0fe39c22c1438bec1811ffe10ec \
|
||||
--hash=sha256:4f1798f56c6f4ba5ac11fa2869e5717926e4470d97a1dd42b4f59219d43b5027 \
|
||||
--hash=sha256:514698d3683189031dcbfdc31e87115992e5ce9e1b19fe5359941323f2df800c \
|
||||
--hash=sha256:53c2c5d7e2024e46776c2d90e6c637d01102126b61aaf5faa5edaf05f8b5722a \
|
||||
--hash=sha256:55d6f4135d914305929fe9e9c44d8b5383a9b3fa1bee3bfcf60ee97e01af07ea \
|
||||
--hash=sha256:5a2bbad3a63007f7e9524d4903ba04fee252557c2acd86f9a3d4f91786695254 \
|
||||
--hash=sha256:5a3e0f56ec19bfd9ad1605572878dd6ff7f01b325f4fc154812ae70d615c3aff \
|
||||
--hash=sha256:5bb9be71d9ee39b4359b832f9569518ec9bc08704194034e79e4958e6bc4d46d \
|
||||
--hash=sha256:60ec798c40a1810d282ee046f61ecac1c5675cb898763d9f08d97d53a5e00a81 \
|
||||
--hash=sha256:6b3cbc4f96ddfa1fb88a78a69dd851369825b7816d9702eee8c4461505ba172e \
|
||||
--hash=sha256:6c7ef4701a96553514b2688e342ef1bf2beae6cfd172d89a76c768292aabf405 \
|
||||
--hash=sha256:7337b06a9f9ed9ea3049f04b76f65819db9b19bb832ee598e97b388eadf25e5f \
|
||||
--hash=sha256:76345f51367fb1f23e08605c6efb18374f669be5b223658fbab6b17627950507 \
|
||||
--hash=sha256:7e35c9bc977760981693e1a7a51493b58ee5a501f9ebb1e547565ee40b6c6208 \
|
||||
--hash=sha256:80cac5cb90ed7b9b72a217a1d6982b15b829cdbd0ee6bc19b93e3a9e47fb0ac9 \
|
||||
--hash=sha256:8af88fe5c7dd68fe1fec6dea8155caa1a47155d219a750ff34049541cf536a5e \
|
||||
--hash=sha256:8fcd721113260ffb5e38bf14a8725b17d431f34209f7d1c7005b667946e630b3 \
|
||||
--hash=sha256:93087a845cdfb35753e539354ac9551bdd2ff528c202a98df0ae46e852bcf021 \
|
||||
--hash=sha256:93935ab27b6eaef4c3e5489aebc84284f0644592f7ab516df60ee1b27eaf5eb3 \
|
||||
--hash=sha256:9bf08e4b6362dd1c08623bbfa2d061e8bac0f1da8fc2007062cfe1dc360a49fa \
|
||||
--hash=sha256:a6ac96da66c35bf789bdcde78a88dc7d56b7907d8379648c54adc1c61594575d \
|
||||
--hash=sha256:ab17b3a8aa754be0de0e397f7b95f13b14e56f077a4c6ae295e3d4afd199b325 \
|
||||
--hash=sha256:b0d106d9314546d69b3df1b5352639aa628530ec3ecef8a98a21942d2a2a64f5 \
|
||||
--hash=sha256:b90397a50ad6332ed3e459c648ac20d182cce24a557354363ad85fc9ea4a17cd \
|
||||
--hash=sha256:bbbdb8fcd5e7062e546eec1ac78c28bb21ae7df54c18f8e4b06e15a18d661a49 \
|
||||
--hash=sha256:bd6f2f81c7b9da99d301c0b16b82044e76fe887086e42e1590ecf520b94dbdac \
|
||||
--hash=sha256:be53e1d5f83de43dc175c87612ecc128d444b38e5c56cb3f807f5a73d6887476 \
|
||||
--hash=sha256:c3f23f403a273900d57de6ee5ca0614c650f7f58563065dad1a4744498960e53 \
|
||||
--hash=sha256:cbe8d03d4efcccdb7fcedecbaa1e1fa02913eaf3a74cb933634a6bc6d2ea9e2a \
|
||||
--hash=sha256:ce17bc317d089f361b33521654c13e30eacfd3d2034fd34e613ca9c51c969686 \
|
||||
--hash=sha256:d918d6f304a309222a784bbd140b85ec5594d97e4dc0e79f590549d28970663a \
|
||||
--hash=sha256:dc61e4f9e37b507069dc7e659ae28bca7adcb04c993c3508214315d12c63f848 \
|
||||
--hash=sha256:e095f9e105af76593b4c183bb0bcbdae64bd913a59ec595732dc108b48730ab5 \
|
||||
--hash=sha256:e2cba180d6451822763eda8364f342435a873bcfb3849cbd82fdeca248ca65eb \
|
||||
--hash=sha256:e49eba8f1e28e7c03648b7a476e1ba05309e087ccdea859fc6dd659564aa8d7e \
|
||||
--hash=sha256:f1341c6abcee1c31277cfe28d34e46196f2143ec3d755e6efe7452126e1f626d \
|
||||
--hash=sha256:f3fbfe82829d8e381426a289b87e59e585278728361db9ce975b88b51f64f410 \
|
||||
--hash=sha256:f50b580fad005a092ed87c5a3a4683459b21d1620497d6a5bccad203bee4c071 \
|
||||
--hash=sha256:f5576e8ae1723ece60d8f93c6710abf784714e99388bcf023ba9ca800bc587f6 \
|
||||
--hash=sha256:f673d835bdb1a60229cc3609a113fd2c9ce3f4a3c75ad4eaed111180c00199d2 \
|
||||
--hash=sha256:f7548ede3ee908cfabc0d068106e303a9a2d811af959cdf6ab85676344cedcda \
|
||||
--hash=sha256:fa8ed556fcc87a4091cf61587ef172fa104323dc89ecc085a618ba7ff8629a8f \
|
||||
--hash=sha256:fefea5cf8cdda9053b962ca8a90216fb0b1d40907dcb6819382b42e483e6e9f6 \
|
||||
--hash=sha256:ff7144d8167e513fe39fbb46bffb4f6f192dfb1f4b0b4e9102e1fd4f212e4747
|
||||
# via -r requirements.in
|
||||
@@ -0,0 +1,162 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Run every check of REPORT.md on the current checkout and save the real output in evidence/.
|
||||
|
||||
python scripts/collect_evidence.py # ~25 min; the stand must NOT be running yet
|
||||
|
||||
Intended for a CLEAN CLONE (fresh volumes): it creates secrets and .env, builds and starts the
|
||||
stand, then runs make check/validate/test/test-persistence/test-failover, the own tests, the
|
||||
level 4 experiments, and records container states before/after every hook. Prints no secrets
|
||||
(hook output is shown by the public tests as suppressed; our scripts print names and states only).
|
||||
On Windows the hooks are the scripts/*.cmd wrappers (Git Bash) and make needs Git's usr/bin.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import datetime as dt
|
||||
import json
|
||||
import os
|
||||
import shutil
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
EVIDENCE = ROOT / "evidence"
|
||||
WINDOWS = os.name == "nt"
|
||||
HOOK = ".cmd" if WINDOWS else ".sh"
|
||||
PYTHON = sys.executable
|
||||
VENV_PY = str(ROOT / ".venv" / ("Scripts/python.exe" if WINDOWS else "bin/python"))
|
||||
# Makefile recipes use $(PYTHON) unquoted: pass interpreters WITHOUT spaces in the path.
|
||||
MAKE_PY = "python" if WINDOWS else "python3"
|
||||
MAKE_VENV = ".venv/Scripts/python.exe" if WINDOWS else ".venv/bin/python"
|
||||
ENV = {**os.environ, "PYTHONUTF8": "1"}
|
||||
if WINDOWS:
|
||||
ENV["PATH"] = ENV["PATH"] + os.pathsep + r"C:\Program Files\Git\usr\bin"
|
||||
|
||||
|
||||
def now() -> str:
|
||||
return dt.datetime.now(dt.timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
||||
|
||||
|
||||
def sh(cmd: list[str], timeout: float = 1800, env: dict | None = None) -> tuple[int, str]:
|
||||
started = time.monotonic()
|
||||
result = subprocess.run(cmd, cwd=ROOT, capture_output=True, text=True, encoding="utf-8", errors="replace",
|
||||
timeout=timeout, env={**ENV, **(env or {})})
|
||||
took = time.monotonic() - started
|
||||
return result.returncode, (result.stdout + result.stderr).rstrip() + f"\n[exit {result.returncode}, {took:.1f} s]\n"
|
||||
|
||||
|
||||
class Log:
|
||||
def __init__(self, name: str, title: str):
|
||||
self.path = EVIDENCE / name
|
||||
self.path.parent.mkdir(parents=True, exist_ok=True)
|
||||
commit = subprocess.run(["git", "log", "-1", "--format=%h %s"], cwd=ROOT, capture_output=True, text=True,
|
||||
encoding="utf-8").stdout.strip()
|
||||
self.text = f"# {title}\n# {now()}, commit {commit}\n"
|
||||
|
||||
def run(self, cmd: list[str], shown: str | None = None, **kwargs) -> int:
|
||||
code, out = sh(cmd, **kwargs)
|
||||
self.text += f"\n$ {shown or ' '.join(cmd)}\n{out}"
|
||||
self.save()
|
||||
print(f" {self.path.name}: {shown or ' '.join(cmd[:4])} -> exit {code}", flush=True)
|
||||
return code
|
||||
|
||||
def note(self, text: str) -> None:
|
||||
self.text += f"\n{text}\n"
|
||||
self.save()
|
||||
|
||||
def save(self) -> None:
|
||||
self.path.write_text(self.text, encoding="utf-8")
|
||||
|
||||
|
||||
def states() -> list[str]:
|
||||
ids = subprocess.run(["docker", "compose", "ps", "-a", "-q"], cwd=ROOT, capture_output=True, text=True).stdout.split()
|
||||
rows = []
|
||||
for item in json.loads(subprocess.run(["docker", "inspect", *ids], capture_output=True, text=True).stdout or "[]"):
|
||||
st = item["State"]
|
||||
rows.append(f"{item['Name'].lstrip('/'):24} host={item['Config']['Hostname']:13} {st['Status']:8} "
|
||||
f"exit={st['ExitCode']:<3} pid={st['Pid']:<6} started={st['StartedAt'][:23]}")
|
||||
return sorted(rows)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
EVIDENCE.mkdir(exist_ok=True)
|
||||
make = shutil.which("make", path=ENV["PATH"]) or "make"
|
||||
# 1. Clean clone -> secrets -> build -> start
|
||||
log = Log("clean-clone.txt", "Clean clone: secrets, config, build, migrations from zero, start, versions")
|
||||
log.run(["git", "status", "--short", "--branch"])
|
||||
log.run([PYTHON, "--version"]); log.run(["docker", "version", "--format", "client {{.Client.Version}} / server {{.Server.Version}}"])
|
||||
log.run(["docker", "compose", "version"]); log.run([make, "--version"], shown="make --version")
|
||||
log.run([PYTHON, "scripts/init_secrets.py"], shown="python scripts/init_secrets.py")
|
||||
if not (ROOT / ".env").exists():
|
||||
shutil.copy(ROOT / ".env.example", ROOT / ".env")
|
||||
log.note("$ cp .env.example .env")
|
||||
log.run(["docker", "compose", "--env-file", ".env", "config", "--quiet"])
|
||||
log.run(["docker", "compose", "--env-file", ".env", "up", "-d", "--build", "--wait"], timeout=1800)
|
||||
log.run(["docker", "compose", "ps", "-a", "--format", "table {{.Service}}\t{{.Image}}\t{{.Status}}"])
|
||||
log.run(["docker", "compose", "logs", "--no-log-prefix", "migrate"])
|
||||
log.run([PYTHON, "scripts/lbctl.py", "replicas"]); log.run([PYTHON, "scripts/lbctl.py", "status"])
|
||||
# venv for validate / own tests
|
||||
if not Path(VENV_PY).exists():
|
||||
log.run([PYTHON, "-m", "venv", ".venv"])
|
||||
log.run([VENV_PY, "-m", "pip", "install", "-q", "--require-hashes", "--no-deps", "-r", "requirements.txt"], timeout=1800)
|
||||
log.run([VENV_PY, "-m", "pip", "install", "-q", "-r", "requirements-test.txt", "-r", "requirements-dev.txt"], timeout=1800)
|
||||
|
||||
# 2. make check / validate / test
|
||||
log = Log("make-check-validate-test.txt", "make check, make validate (venv), make test")
|
||||
log.run([make, "check", f"PYTHON={MAKE_PY}"], shown="make check")
|
||||
log.run([make, "validate", f"PYTHON={MAKE_VENV}"], shown=f"make validate PYTHON={MAKE_VENV}")
|
||||
log.run([make, "test", f"PYTHON={MAKE_PY}"], shown="make test")
|
||||
|
||||
# 3. persistence with states before/after
|
||||
log = Log("test-persistence.txt", "make test-persistence: restart of LB, all API replicas, PostgreSQL, Valkey")
|
||||
log.note("Containers BEFORE:\n" + "\n".join(states()))
|
||||
log.run([make, "test-persistence", f"ACTION_SCRIPT=scripts/restart{HOOK}", f"PYTHON={MAKE_PY}"],
|
||||
shown=f"make test-persistence ACTION_SCRIPT=scripts/restart{HOOK}")
|
||||
log.note("Containers AFTER (new pid/started for lb, api-*, postgres, valkey; volumes kept):\n" + "\n".join(states()))
|
||||
log.run([PYTHON, "scripts/lbctl.py", "wait-up", "2"])
|
||||
|
||||
# 4. failover: until BOTH replicas have been the stopped one (each run is a real public test)
|
||||
log = Log("test-failover.txt", "make test-failover with scripts/stop-one: which container is really stopped, restore, repeat")
|
||||
stopped: set[str] = set()
|
||||
import urllib.request
|
||||
for attempt in range(1, 7):
|
||||
# The public test stops the replica that served its first POST; its request sequence is
|
||||
# fixed, so with round robin it hits the same replica every time. Shift the rotation by
|
||||
# (attempt - 1) extra requests so that the other replica becomes the target as well.
|
||||
for _ in range(attempt - 1):
|
||||
urllib.request.urlopen("http://127.0.0.1:8080/health/live", timeout=5).read()
|
||||
before = {r for r in states() if "-api-" in r and " running " in r}
|
||||
log.note(f"--- run {attempt}; running API replicas BEFORE:\n" + "\n".join(sorted(before)))
|
||||
log.run([make, "test-failover", f"ACTION_SCRIPT=scripts/stop-one{HOOK}", f"PYTHON={MAKE_PY}"],
|
||||
shown=f"make test-failover ACTION_SCRIPT=scripts/stop-one{HOOK}")
|
||||
after = states()
|
||||
gone = [r for r in after if "-api-" in r and " exited " in r]
|
||||
log.note("AFTER (the stopped replica, left stopped by the hook):\n" + "\n".join(gone))
|
||||
log.run([PYTHON, "scripts/lbctl.py", "status"])
|
||||
stopped.update(r.split()[0] for r in gone)
|
||||
log.run(["sh" if not WINDOWS else r"C:\Program Files\Git\bin\bash.exe", "scripts/restore-api.sh"],
|
||||
shown="sh scripts/restore-api.sh")
|
||||
log.run([PYTHON, "scripts/lbctl.py", "wait-up", "2"])
|
||||
if len(stopped) >= 2:
|
||||
break
|
||||
log.note(f"Replicas stopped by the public test across runs: {sorted(stopped)}")
|
||||
|
||||
# 5. own tests
|
||||
log = Log("student-tests.txt", "Own tests: unit + stand (pytest -v); failures restored in finally")
|
||||
log.run([VENV_PY, "-m", "pytest", "student-tests", "-v", "-p", "no:cacheprovider"], shown="python -m pytest student-tests -v",
|
||||
timeout=3600)
|
||||
|
||||
# 6. level 4 experiments and the public load script
|
||||
log = Log("level4-runs.txt", "Level 4: failover series, rolling update under load, public load.py")
|
||||
log.run([PYTHON, "bench/failover_series.py", "--output", "evidence/failover"], timeout=1800)
|
||||
log.run([PYTHON, "bench/rolling_under_load.py", "--output", "evidence/rolling-update"], timeout=1800)
|
||||
log.run([PYTHON, "scripts/lbctl.py", "wait-up", "2"])
|
||||
log.run([PYTHON, "tests/load.py", "--duration", "30", "--concurrency", "4", "--output", "evidence/load.json"])
|
||||
log.run(["docker", "compose", "ps", "-a", "--format", "table {{.Service}}\t{{.Status}}"])
|
||||
print("done:", EVIDENCE)
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1,114 @@
|
||||
#!/usr/bin/env python3
|
||||
"""EXPLAIN (ANALYZE, BUFFERS) of the history query on >= 10 000 messages (level 4).
|
||||
|
||||
python scripts/explain_history.py --output evidence/explain-history.txt
|
||||
|
||||
Creates (once, idempotent) a synthetic chat with 20 000 messages plus 30 000 messages in
|
||||
30 other chats, then explains the EXACT statements of src/messenger/repo.py:
|
||||
1. first page (HISTORY_FIRST)
|
||||
2. late page via cursor (HISTORY_AFTER) — generic plan forced, as a prepared statement
|
||||
reused by the driver would get it
|
||||
3. the rejected variant `($2 IS NULL OR (created_at, id) > ...)` — generic plan
|
||||
4. the rejected OFFSET pagination for the same late page
|
||||
Runs psql inside the postgres container; no database port is published.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import subprocess
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
sys.path.insert(0, str(ROOT / "src"))
|
||||
from messenger.repo import HISTORY_AFTER, HISTORY_FIRST # noqa: E402 (the app's real SQL)
|
||||
|
||||
CHAT = "00000000-0000-4000-8000-00000000e002"
|
||||
USER = "00000000-0000-4000-8000-00000000e001"
|
||||
TARGET = 20_000
|
||||
LATE = 19_950 # cursor position: the 50 newest messages remain
|
||||
|
||||
SEED = f"""
|
||||
\\set ON_ERROR_STOP on
|
||||
BEGIN;
|
||||
INSERT INTO users (id, username, display_name, password_hash)
|
||||
VALUES ('{USER}', 'explain_seed', convert_to('Explain seed', 'UTF8'), 'not-a-login') ON CONFLICT DO NOTHING;
|
||||
INSERT INTO chats (id, title, created_by, created_at)
|
||||
SELECT ('00000000-0000-4000-8000-0000000' || lpad(to_hex(n), 5, '0'))::uuid, convert_to('explain ' || n, 'UTF8'),
|
||||
'{USER}', timestamptz '2026-01-01 00:00:00+00'
|
||||
FROM generate_series(57346::int, 57376::int) AS n -- e002 .. e020: target + 30 noise chats
|
||||
ON CONFLICT DO NOTHING;
|
||||
INSERT INTO chat_members (chat_id, user_id, position, chat_created_at)
|
||||
SELECT id, '{USER}', 0, created_at FROM chats WHERE created_by = '{USER}' ON CONFLICT DO NOTHING;
|
||||
INSERT INTO messages (id, chat_id, sender_id, text, created_at)
|
||||
SELECT gen_random_uuid(), '{CHAT}', '{USER}', convert_to('synthetic message ' || g, 'UTF8'),
|
||||
timestamptz '2026-01-01 00:00:00+00' + g * interval '1 second'
|
||||
FROM generate_series(1::int, {TARGET}::int) AS g
|
||||
WHERE NOT EXISTS (SELECT 1 FROM messages WHERE chat_id = '{CHAT}');
|
||||
INSERT INTO messages (id, chat_id, sender_id, text, created_at)
|
||||
SELECT gen_random_uuid(), c.id, '{USER}', convert_to('noise ' || g, 'UTF8'),
|
||||
timestamptz '2026-01-01 00:00:00+00' + g * interval '1 second'
|
||||
FROM chats c CROSS JOIN generate_series(1::int, 1000::int) AS g
|
||||
WHERE c.created_by = '{USER}' AND c.id <> '{CHAT}'
|
||||
AND NOT EXISTS (SELECT 1 FROM messages m WHERE m.chat_id = c.id);
|
||||
COMMIT;
|
||||
ANALYZE messages;
|
||||
"""
|
||||
|
||||
|
||||
def pg(sql_param: str) -> str:
|
||||
"""Turn psycopg %s placeholders into $1..$n for a PREPARE statement."""
|
||||
parts = sql_param.split("%s")
|
||||
return "".join(p + (f"${i + 1}" if i < len(parts) - 1 else "") for i, p in enumerate(parts))
|
||||
|
||||
|
||||
EXPLAIN = f"""
|
||||
\\set ON_ERROR_STOP on
|
||||
\\pset pager off
|
||||
SELECT count(*) AS messages_total, count(*) FILTER (WHERE chat_id = '{CHAT}') AS messages_in_target_chat FROM messages;
|
||||
SELECT created_at AS cur_ts, id AS cur_id FROM messages WHERE chat_id = '{CHAT}'
|
||||
ORDER BY created_at, id OFFSET {LATE - 1} LIMIT 1 \\gset
|
||||
SET plan_cache_mode = force_generic_plan;
|
||||
PREPARE first_page(uuid, int) AS {pg(HISTORY_FIRST)};
|
||||
PREPARE after_page(uuid, timestamptz, uuid, int) AS {pg(HISTORY_AFTER)};
|
||||
PREPARE or_page(uuid, timestamptz, uuid, int) AS
|
||||
SELECT id, sender_id, text, created_at FROM messages
|
||||
WHERE chat_id = $1 AND ($2 IS NULL OR (created_at, id) > ($2, $3)) ORDER BY created_at, id LIMIT $4;
|
||||
PREPARE offset_page(uuid, int, int) AS
|
||||
SELECT id, sender_id, text, created_at FROM messages WHERE chat_id = $1 ORDER BY created_at, id OFFSET $2 LIMIT $3;
|
||||
\\echo '=== 1. first page, app query HISTORY_FIRST, generic plan'
|
||||
EXPLAIN (ANALYZE, BUFFERS, SUMMARY ON) EXECUTE first_page('{CHAT}', 51);
|
||||
\\echo '=== 2. late page (cursor at {LATE}), app query HISTORY_AFTER, generic plan'
|
||||
EXPLAIN (ANALYZE, BUFFERS, SUMMARY ON) EXECUTE after_page('{CHAT}', :'cur_ts', :'cur_id', 51);
|
||||
\\echo '=== 3. REJECTED: one query with ($2 IS NULL OR ...), same late page, generic plan'
|
||||
EXPLAIN (ANALYZE, BUFFERS, SUMMARY ON) EXECUTE or_page('{CHAT}', :'cur_ts', :'cur_id', 51);
|
||||
\\echo '=== 4. REJECTED: OFFSET pagination, same late page'
|
||||
EXPLAIN (ANALYZE, BUFFERS, SUMMARY ON) EXECUTE offset_page('{CHAT}', {LATE}, 51);
|
||||
"""
|
||||
|
||||
|
||||
def psql(script: str) -> str:
|
||||
result = subprocess.run(
|
||||
["docker", "compose", "exec", "-T", "postgres", "psql", "-U", "messenger", "-d", "messenger", "-X", "-q", "-f", "-"],
|
||||
cwd=ROOT, input=script, capture_output=True, text=True, encoding="utf-8", check=True)
|
||||
return result.stdout
|
||||
|
||||
|
||||
def run() -> str:
|
||||
psql(SEED)
|
||||
return psql(EXPLAIN)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description=__doc__.splitlines()[0])
|
||||
parser.add_argument("--output", default=None)
|
||||
args = parser.parse_args()
|
||||
text = run()
|
||||
print(text)
|
||||
if args.output:
|
||||
Path(args.output).write_text(text, encoding="utf-8", newline="\n")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
@@ -0,0 +1,47 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Create local synthetic credentials for the stand in ./secrets (ignored by Git).
|
||||
|
||||
python scripts/init_secrets.py
|
||||
|
||||
Idempotent: existing files are kept, so re-running never changes passwords of a
|
||||
database that already has data. Prints only file names, never values.
|
||||
Delete ./secrets together with the volumes (`docker compose down -v`) to start over.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
import secrets
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
SECRETS = ROOT / "secrets"
|
||||
|
||||
|
||||
def write(name: str, value: str) -> None:
|
||||
path = SECRETS / name
|
||||
if path.exists():
|
||||
print(f"keep secrets/{name}")
|
||||
return
|
||||
path.write_text(value + "\n", encoding="utf-8", newline="\n")
|
||||
# Readable by the container users (postgres, valkey, app uid 10001) through the
|
||||
# compose bind mount. Local synthetic lab credentials only.
|
||||
os.chmod(path, 0o644)
|
||||
print(f"created secrets/{name}")
|
||||
|
||||
|
||||
def read(name: str) -> str:
|
||||
return (SECRETS / name).read_text(encoding="utf-8").strip()
|
||||
|
||||
|
||||
def main() -> int:
|
||||
SECRETS.mkdir(exist_ok=True)
|
||||
write("postgres_password", secrets.token_urlsafe(24)) # [A-Za-z0-9_-]: safe inside URLs
|
||||
write("valkey_password", secrets.token_urlsafe(24))
|
||||
write("database_url", f"postgresql://messenger:{read('postgres_password')}@postgres:5432/messenger")
|
||||
write("cache_url", f"redis://:{read('valkey_password')}@valkey:6379/0")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
@@ -0,0 +1,125 @@
|
||||
#!/usr/bin/env python3
|
||||
"""What the load balancer thinks right now, mapped to real containers.
|
||||
|
||||
python scripts/lbctl.py status # table: HAProxy slot, IP, UP/DOWN/MAINT, container, X-Instance-Id
|
||||
python scripts/lbctl.py wait-up 2 # wait until exactly 2 API replicas are UP in HAProxy
|
||||
python scripts/lbctl.py replicas # api containers of this compose project: name, hostname, state
|
||||
|
||||
Reads HAProxy's runtime API (`show servers state api`) through `docker compose exec lb socat`,
|
||||
so no admin port is published. Container hostname = X-Instance-Id of that replica.
|
||||
Prints no secrets. Used by student-tests and bench scripts as a library as well.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
|
||||
# srv_op_state of HAProxy (src/server.h): 0 stopped, 1 starting, 2 running, 3 stopping.
|
||||
OP_STATE = {0: "DOWN", 1: "STARTING", 2: "UP", 3: "STOPPING"}
|
||||
ADMIN_RMAINT = 0x20 # slot without a resolved address (no running container behind it)
|
||||
|
||||
|
||||
def run(*args: str, timeout: float = 60) -> str:
|
||||
result = subprocess.run(list(args), cwd=ROOT, capture_output=True, text=True, encoding="utf-8",
|
||||
timeout=timeout)
|
||||
if result.returncode != 0:
|
||||
raise RuntimeError(f"{' '.join(args[:3])}... failed: {result.stderr.strip()[:300]}")
|
||||
return result.stdout
|
||||
|
||||
|
||||
def replicas(all_states: bool = True) -> list[dict]:
|
||||
"""API containers of this compose project (running and stopped)."""
|
||||
ids = run("docker", "compose", "ps", "-q", *(["-a"] if all_states else []), "api").split()
|
||||
if not ids:
|
||||
return []
|
||||
data = json.loads(run("docker", "inspect", *ids))
|
||||
result = []
|
||||
for item in data:
|
||||
networks = item["NetworkSettings"]["Networks"] or {}
|
||||
result.append({
|
||||
"id": item["Id"][:12],
|
||||
"name": item["Name"].lstrip("/"),
|
||||
"hostname": item["Config"]["Hostname"],
|
||||
"state": item["State"]["Status"],
|
||||
"health": (item["State"].get("Health") or {}).get("Status", "-"),
|
||||
"exit_code": item["State"]["ExitCode"],
|
||||
"started_at": item["State"]["StartedAt"],
|
||||
"pid": item["State"]["Pid"],
|
||||
"ips": sorted(n["IPAddress"] for n in networks.values() if n.get("IPAddress")),
|
||||
# compose network name without the project prefix -> IP, e.g. {"edge": ..., "data": ...}
|
||||
"networks": {name.split("_", 1)[-1]: n["IPAddress"] for name, n in networks.items() if n.get("IPAddress")},
|
||||
})
|
||||
return sorted(result, key=lambda r: r["name"])
|
||||
|
||||
|
||||
def lb_servers() -> list[dict]:
|
||||
raw = run("docker", "compose", "exec", "-T", "lb", "sh", "-c",
|
||||
'echo "show servers state api" | socat stdio /var/lib/haproxy/admin.sock')
|
||||
by_ip = {ip: r for r in replicas() for ip in r["ips"]}
|
||||
servers = []
|
||||
for line in raw.splitlines():
|
||||
parts = line.split()
|
||||
if len(parts) < 7 or not parts[0].isdigit():
|
||||
continue
|
||||
name, addr, op, admin = parts[3], parts[4], int(parts[5]), int(parts[6])
|
||||
if admin & ADMIN_RMAINT or addr == "-":
|
||||
state = "MAINT"
|
||||
else:
|
||||
state = OP_STATE.get(op, str(op))
|
||||
container = by_ip.get(addr)
|
||||
servers.append({"slot": name, "addr": addr, "state": state,
|
||||
"container": container["name"] if container else None,
|
||||
"instance_id": container["hostname"] if container else None})
|
||||
return servers
|
||||
|
||||
|
||||
def up_instances() -> set[str]:
|
||||
return {s["instance_id"] or s["addr"] for s in lb_servers() if s["state"] == "UP"}
|
||||
|
||||
|
||||
def wait_up(count: int, timeout: float = 60) -> float:
|
||||
"""Seconds until exactly `count` servers are UP (raises TimeoutError otherwise)."""
|
||||
started = time.monotonic()
|
||||
while True:
|
||||
try:
|
||||
if len(up_instances()) == count:
|
||||
return time.monotonic() - started
|
||||
except RuntimeError:
|
||||
pass # lb restarting
|
||||
if time.monotonic() - started > timeout:
|
||||
raise TimeoutError(f"HAProxy does not report {count} UP API replicas after {timeout:.0f} s")
|
||||
time.sleep(0.25)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
|
||||
sub = parser.add_subparsers(dest="command", required=True)
|
||||
sub.add_parser("status")
|
||||
sub.add_parser("replicas")
|
||||
wait = sub.add_parser("wait-up")
|
||||
wait.add_argument("count", type=int)
|
||||
wait.add_argument("--timeout", type=float, default=60)
|
||||
args = parser.parse_args()
|
||||
if args.command == "status":
|
||||
for s in lb_servers():
|
||||
if s["state"] != "MAINT" or s["container"]:
|
||||
print(f"{s['slot']:6} {s['addr']:15} {s['state']:8} {s['container'] or '-':24} X-Instance-Id={s['instance_id'] or '-'}")
|
||||
print(f"(other slots: MAINT, no container)")
|
||||
elif args.command == "replicas":
|
||||
for r in replicas():
|
||||
print(f"{r['name']:24} X-Instance-Id={r['hostname']} {r['state']:8} health={r['health']:9} "
|
||||
f"pid={r['pid']} started={r['started_at'][:23]} ip={','.join(r['ips']) or '-'}")
|
||||
else:
|
||||
took = wait_up(args.count, args.timeout)
|
||||
print(f"{args.count} API replica(s) UP in HAProxy after {took:.1f} s: {sorted(up_instances())}")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
@@ -0,0 +1,6 @@
|
||||
@echo off
|
||||
rem Windows wrapper: tests/resilience.py starts the hook as a program, and Windows
|
||||
rem cannot execute a .sh file directly. Runs the same scripts\restart.sh with Git Bash.
|
||||
rem python tests/resilience.py --mode persistence --action-script scripts/restart.cmd
|
||||
"%ProgramFiles%\Git\bin\bash.exe" "%~dp0restart.sh"
|
||||
exit /b %ERRORLEVEL%
|
||||
Executable
+17
@@ -0,0 +1,17 @@
|
||||
#!/bin/sh
|
||||
# Restart hook for `make test-persistence ACTION_SCRIPT=scripts/restart.sh`.
|
||||
# Restarts EVERY process of the stand: LB, all API replicas, PostgreSQL and Valkey.
|
||||
# Volumes, schema and data are preserved: no `down -v`, no FLUSH, no re-seed.
|
||||
# Output is suppressed by the test; run it by hand to see the steps.
|
||||
set -eu
|
||||
cd "$(dirname "$0")/.."
|
||||
|
||||
# 1. Entry point and API first: no request is served while the stores restart underneath.
|
||||
docker compose stop lb api
|
||||
# 2. Stores: clean shutdown (PostgreSQL checkpoint, Valkey AOF fsync), then start again.
|
||||
docker compose restart postgres valkey
|
||||
# 3. API replicas (wait for healthy stores via depends_on), then the LB. HAProxy resolves
|
||||
# the new replica addresses through Docker DNS by itself - no stale backend list.
|
||||
docker compose start api
|
||||
docker compose start lb
|
||||
echo "restart.sh: lb, api (all replicas), postgres and valkey restarted with volumes preserved"
|
||||
Executable
+8
@@ -0,0 +1,8 @@
|
||||
#!/bin/sh
|
||||
# Start every stopped API replica again (after scripts/stop-one.sh) and wait until Docker
|
||||
# reports them healthy. The LB picks them up by itself (DNS re-resolution + checks, ~2-3 s);
|
||||
# `python scripts/lbctl.py wait-up 2` waits for that explicitly.
|
||||
set -eu
|
||||
cd "$(dirname "$0")/.."
|
||||
docker compose up -d --no-recreate --no-deps --wait api
|
||||
docker compose ps api --format '{{.Name}} {{.Status}}'
|
||||
Executable
+37
@@ -0,0 +1,37 @@
|
||||
#!/bin/sh
|
||||
# Rolling update of the API without downtime (Compose has no built-in one for a scaled service).
|
||||
# 1. build the new image from the working tree;
|
||||
# 2. run migrations ONCE with the new image (they must be additive / compatible with the
|
||||
# running old version: expand first, contract in a later release);
|
||||
# 3. for every OLD replica: start one NEW replica (scale N+1, old ones untouched), wait until
|
||||
# Docker says healthy AND HAProxy has N+1 replicas UP, then stop the old one gracefully
|
||||
# (SIGTERM -> readiness 503 -> HAProxy DOWN -> drain -> exit) and remove it.
|
||||
# At every moment at least N replicas serve traffic. Stops on the first error (set -e);
|
||||
# the old replicas that were not replaced yet keep running.
|
||||
set -eu
|
||||
cd "$(dirname "$0")/.."
|
||||
PY="${PYTHON:-$(command -v python3 || command -v python)}"
|
||||
# New image = new revision label (git commit + UTC time), visible in docker inspect and logs.
|
||||
BUILD_REVISION="${BUILD_REVISION:-$(git rev-parse --short HEAD 2>/dev/null || echo dev)-$(date -u +%Y%m%dT%H%M%SZ)}"
|
||||
export BUILD_REVISION
|
||||
|
||||
replicas="$(docker compose ps -q api | wc -l | tr -d ' ')"
|
||||
[ "$replicas" -ge 1 ] || { echo "rolling-update: no running api replica" >&2; exit 2; }
|
||||
old="$(docker compose ps -q api)"
|
||||
echo "rolling-update: $replicas replica(s) to replace, new revision $BUILD_REVISION"
|
||||
|
||||
docker compose build api
|
||||
docker compose run --rm --no-deps migrate
|
||||
|
||||
for id in $old; do
|
||||
name="$(docker inspect -f '{{.Name}}' "$id" | tr -d '/')"
|
||||
docker compose up -d --no-deps --no-recreate --wait --scale "api=$((replicas + 1))" api
|
||||
"$PY" scripts/lbctl.py wait-up "$((replicas + 1))" --timeout 60
|
||||
docker stop --time 15 "$id" >/dev/null
|
||||
docker rm "$id" >/dev/null
|
||||
"$PY" scripts/lbctl.py wait-up "$replicas" --timeout 60
|
||||
echo "rolling-update: replaced $name"
|
||||
done
|
||||
for id in $(docker compose ps -q api); do
|
||||
docker inspect -f '{{.Name}} revision={{index .Config.Labels "org.opencontainers.image.revision"}} {{.State.Status}}' "$id"
|
||||
done
|
||||
@@ -0,0 +1,7 @@
|
||||
@echo off
|
||||
rem Windows wrapper: tests/resilience.py starts the hook as a program, and Windows
|
||||
rem cannot execute a .sh file directly. Runs the same scripts\stop-one.sh with Git Bash;
|
||||
rem TARGET_INSTANCE_ID and STOP_MODE are inherited from the environment.
|
||||
rem python tests/resilience.py --mode failover --action-script scripts/stop-one.cmd
|
||||
"%ProgramFiles%\Git\bin\bash.exe" "%~dp0stop-one.sh"
|
||||
exit /b %ERRORLEVEL%
|
||||
Executable
+35
@@ -0,0 +1,35 @@
|
||||
#!/bin/sh
|
||||
# Failover hook for `make test-failover ACTION_SCRIPT=scripts/stop-one.sh`.
|
||||
# Stops exactly ONE API replica: the container whose X-Instance-Id (= container hostname,
|
||||
# see src/messenger/config.py) is in TARGET_INSTANCE_ID. DB, Valkey, LB and volumes are
|
||||
# not touched. The replica is LEFT STOPPED (`docker stop` = no restart by the policy);
|
||||
# restore it explicitly afterwards: scripts/restore-api.sh.
|
||||
# STOP_MODE=stop (default) SIGTERM: readiness 503, 3 s drain, graceful exit (~4 s)
|
||||
# STOP_MODE=kill SIGKILL at once (`docker stop -t 0`): crash of the process
|
||||
# Returns within 10 s. The test hides the output; run by hand to see it (no secrets printed).
|
||||
set -eu
|
||||
cd "$(dirname "$0")/.."
|
||||
: "${TARGET_INSTANCE_ID:?test provides target API instance alias}"
|
||||
case "$TARGET_INSTANCE_ID" in
|
||||
*[!A-Za-z0-9._-]*) echo "stop-one: invalid TARGET_INSTANCE_ID" >&2; exit 2 ;;
|
||||
esac
|
||||
mode="${STOP_MODE:-stop}"
|
||||
|
||||
target=""
|
||||
# Only RUNNING containers of THIS compose project's service `api` are candidates.
|
||||
for id in $(docker compose ps -q api); do
|
||||
if [ "$(docker inspect -f '{{.Config.Hostname}}' "$id")" = "$TARGET_INSTANCE_ID" ]; then
|
||||
target="$id"
|
||||
fi
|
||||
done
|
||||
if [ -z "$target" ]; then
|
||||
echo "stop-one: no running api replica with X-Instance-Id=$TARGET_INSTANCE_ID" >&2
|
||||
exit 3
|
||||
fi
|
||||
name="$(docker inspect -f '{{.Name}}' "$target")"
|
||||
case "$mode" in
|
||||
stop) docker stop --time 15 "$target" >/dev/null ;;
|
||||
kill) docker stop --time 0 "$target" >/dev/null ;;
|
||||
*) echo "stop-one: STOP_MODE must be stop or kill" >&2; exit 2 ;;
|
||||
esac
|
||||
echo "stop-one: $mode $name (X-Instance-Id=$TARGET_INSTANCE_ID) -> $(docker inspect -f '{{.State.Status}} exit={{.State.ExitCode}}' "$target")"
|
||||
@@ -0,0 +1,3 @@
|
||||
"""In-memory messenger for lab 1 (MTUSI, «Высоконагруженные приложения»)."""
|
||||
|
||||
__version__ = "1.0.0"
|
||||
@@ -0,0 +1,88 @@
|
||||
"""Entry point: `python -m messenger`. The HTTP server is the main process (PID 1 in Docker).
|
||||
|
||||
Shutdown sequence on SIGTERM/SIGINT (lab 3: the replica stands behind HAProxy):
|
||||
1. readiness flips to 503 (`app.state.draining = True`), requests are still served;
|
||||
2. after DRAIN_SECONDS (compose: 3 s > HAProxy detection 2 x 1 s) the load balancer has
|
||||
marked this replica DOWN and sends nothing new here;
|
||||
3. uvicorn closes the listening socket; in-flight requests get up to
|
||||
SHUTDOWN_TIMEOUT_SECONDS (compose: 8 s) to finish;
|
||||
4. the lifespan closes the DB pool and the Valkey client; the process exits with code 0.
|
||||
DRAIN + SHUTDOWN (11 s) < compose stop_grace_period (15 s), so Docker never needs SIGKILL.
|
||||
A second signal during the drain skips the rest of it. Data stays in PostgreSQL and Valkey.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import os
|
||||
import sys
|
||||
import threading
|
||||
from types import FrameType
|
||||
|
||||
import uvicorn
|
||||
|
||||
from . import __version__, logging_setup
|
||||
from .app import create_app
|
||||
from .config import ConfigError, Settings
|
||||
|
||||
log = logging.getLogger("messenger")
|
||||
|
||||
|
||||
class GracefulServer(uvicorn.Server):
|
||||
"""uvicorn.Server that drains (readiness 503, still serving) before the normal graceful stop."""
|
||||
|
||||
def __init__(self, config: uvicorn.Config, drain_seconds: float):
|
||||
super().__init__(config)
|
||||
self.drain_seconds = drain_seconds
|
||||
self.drain_timer: threading.Timer | None = None
|
||||
|
||||
def handle_exit(self, sig: int, frame: FrameType | None) -> None:
|
||||
self.config.app.state.draining = True # config.app is our Starlette instance
|
||||
if self.drain_timer is None and self.drain_seconds > 0 and not self.should_exit:
|
||||
log.info("drain_started", extra={"fields": {"signal": sig, "drain_s": self.drain_seconds,
|
||||
"timeout_s": self.config.timeout_graceful_shutdown}})
|
||||
# uvicorn polls should_exit in its main loop, so setting it from a timer thread is safe.
|
||||
self.drain_timer = threading.Timer(self.drain_seconds, self._stop, args=(sig, frame))
|
||||
self.drain_timer.daemon = True
|
||||
self.drain_timer.start()
|
||||
return
|
||||
if self.drain_timer is not None:
|
||||
self.drain_timer.cancel() # second signal: stop draining now
|
||||
self._stop(sig, frame)
|
||||
|
||||
def _stop(self, sig: int, frame: FrameType | None) -> None:
|
||||
if not self.should_exit:
|
||||
log.info("shutdown_requested", extra={"fields": {"signal": sig}})
|
||||
super().handle_exit(sig, frame)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
try:
|
||||
settings = Settings.from_env()
|
||||
except ConfigError as exc:
|
||||
print(f"configuration error: {exc}", file=sys.stderr)
|
||||
return 2
|
||||
logging_setup.configure(settings.log_level)
|
||||
app = create_app(settings)
|
||||
config = uvicorn.Config(
|
||||
app,
|
||||
host=settings.host,
|
||||
port=settings.port,
|
||||
log_config=None,
|
||||
access_log=False,
|
||||
server_header=False,
|
||||
proxy_headers=False,
|
||||
lifespan="on",
|
||||
timeout_graceful_shutdown=settings.shutdown_timeout,
|
||||
)
|
||||
server = GracefulServer(config, settings.drain_seconds)
|
||||
log.info("starting", extra={"fields": {
|
||||
"version": __version__, "revision": os.environ.get("APP_REVISION", "dev"), "env": settings.env, "host": settings.host, "port": settings.port,
|
||||
"python": sys.version.split()[0], "storage": "postgresql", "instance_id": settings.instance_id,
|
||||
}})
|
||||
server.run()
|
||||
log.info("stopped")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
@@ -0,0 +1,423 @@
|
||||
"""HTTP layer: Starlette routes, request-id/logging middleware and error mapping."""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import contextlib
|
||||
import hashlib
|
||||
import json
|
||||
import logging
|
||||
import re
|
||||
import time
|
||||
import uuid
|
||||
from typing import Any
|
||||
|
||||
from starlette.applications import Starlette
|
||||
from starlette.exceptions import HTTPException
|
||||
from starlette.requests import Request
|
||||
from starlette.responses import JSONResponse, Response
|
||||
from starlette.routing import Route
|
||||
from starlette.types import ASGIApp, Message as AsgiMessage, Receive, Scope, Send
|
||||
|
||||
from . import authhdr, cursor, docs, passwords, repo, validation
|
||||
from .config import Settings
|
||||
from .db import Database
|
||||
from .errors import BASIC_CHALLENGE, ApiError, InvalidRequest, NotFound, PayloadTooLarge, Unauthorized, Unavailable
|
||||
from .sessions import SessionStore, build_store
|
||||
from .timefmt import rfc3339
|
||||
|
||||
log = logging.getLogger("messenger")
|
||||
access_log = logging.getLogger("messenger.access")
|
||||
|
||||
REQUEST_ID_RE = re.compile(r"[A-Za-z0-9._:-]{1,128}")
|
||||
|
||||
|
||||
# --------------------------------------------------------------------------
|
||||
# Responses
|
||||
# --------------------------------------------------------------------------
|
||||
def error_response(request_id: str, status: int, code: str, message: str, headers: dict[str, str] | None = None) -> JSONResponse:
|
||||
body = {"error": {"code": code, "message": message, "request_id": request_id}}
|
||||
return JSONResponse(body, status_code=status, headers={"X-Request-Id": request_id, **(headers or {})})
|
||||
|
||||
|
||||
async def api_error_handler(request: Request, exc: Exception) -> JSONResponse:
|
||||
assert isinstance(exc, ApiError)
|
||||
return error_response(request.state.request_id, exc.status, exc.code, exc.message, exc.headers)
|
||||
|
||||
|
||||
async def http_exception_handler(request: Request, exc: Exception) -> JSONResponse:
|
||||
"""Unknown routes and wrong methods also answer in the contract error format."""
|
||||
assert isinstance(exc, HTTPException)
|
||||
if exc.status_code == 404:
|
||||
return error_response(request.state.request_id, 404, "not_found", "Объект не найден")
|
||||
if exc.status_code == 405:
|
||||
return error_response(request.state.request_id, 405, "invalid_request", "Метод не поддерживается", exc.headers)
|
||||
return error_response(request.state.request_id, exc.status_code, "invalid_request", "Неверный запрос")
|
||||
|
||||
|
||||
# --------------------------------------------------------------------------
|
||||
# Middleware (pure ASGI, so it also covers 404/405 and unexpected exceptions)
|
||||
# --------------------------------------------------------------------------
|
||||
class RequestContextMiddleware:
|
||||
"""Adds X-Request-Id and X-Instance-Id to EVERY response (errors, health and crash 500/503
|
||||
included), writes one access-log line, turns crashes into 500."""
|
||||
|
||||
def __init__(self, app: ASGIApp, instance_id: str):
|
||||
self.app = app
|
||||
self.instance_id = instance_id.encode("ascii")
|
||||
|
||||
async def __call__(self, scope: Scope, receive: Receive, send: Send) -> None:
|
||||
if scope["type"] != "http":
|
||||
await self.app(scope, receive, send)
|
||||
return
|
||||
incoming = dict(scope["headers"]).get(b"x-request-id", b"").decode("latin-1")
|
||||
request_id = incoming if REQUEST_ID_RE.fullmatch(incoming) else uuid.uuid4().hex
|
||||
scope.setdefault("state", {})["request_id"] = request_id
|
||||
status = 500
|
||||
started = False
|
||||
begin = time.perf_counter()
|
||||
|
||||
async def send_with_id(message: AsgiMessage) -> None:
|
||||
nonlocal status, started
|
||||
if message["type"] == "http.response.start":
|
||||
started = True
|
||||
status = message["status"]
|
||||
headers = [(k, v) for k, v in message.get("headers", [])
|
||||
if k.lower() not in (b"x-request-id", b"x-instance-id")]
|
||||
headers.append((b"x-request-id", request_id.encode("latin-1")))
|
||||
headers.append((b"x-instance-id", self.instance_id))
|
||||
message = {**message, "headers": headers}
|
||||
await send(message)
|
||||
|
||||
try:
|
||||
await self.app(scope, receive, send_with_id)
|
||||
except asyncio.CancelledError:
|
||||
# uvicorn cancels requests still running when SHUTDOWN_TIMEOUT_SECONDS expires:
|
||||
# answer 503 (retryable) instead of a generic 500, then let the cancellation proceed.
|
||||
if not started:
|
||||
status = 503
|
||||
response = error_response(request_id, 503, "unavailable", "Сервер завершает работу")
|
||||
await response(scope, receive, send_with_id)
|
||||
raise
|
||||
except Exception:
|
||||
log.exception("unhandled_error", extra={"fields": {"request_id": request_id}})
|
||||
if not started:
|
||||
status = 500
|
||||
response = error_response(request_id, 500, "internal_error", "Внутренняя ошибка сервера")
|
||||
await response(scope, receive, send_with_id)
|
||||
finally:
|
||||
access_log.info("request", extra={"fields": {
|
||||
"request_id": request_id,
|
||||
"method": scope["method"],
|
||||
"path": scope["path"], # ids only; query (cursor) and bodies are never logged
|
||||
"status": status,
|
||||
"instance_id": self.instance_id.decode("ascii"),
|
||||
"duration_ms": round((time.perf_counter() - begin) * 1000, 3),
|
||||
}})
|
||||
|
||||
|
||||
# --------------------------------------------------------------------------
|
||||
# Request helpers
|
||||
# --------------------------------------------------------------------------
|
||||
def db_of(request: Request) -> Database:
|
||||
return request.app.state.db
|
||||
|
||||
|
||||
def sessions_of(request: Request) -> SessionStore:
|
||||
return request.app.state.sessions
|
||||
|
||||
|
||||
async def authenticate(request: Request) -> str:
|
||||
"""User id of a valid Bearer session, otherwise 401. X-User-Id is ignored since lab 2.
|
||||
|
||||
Store failures are 503 (Unavailable), never 401 and never anonymous access.
|
||||
"""
|
||||
token = authhdr.parse_bearer(request.headers.get("authorization"))
|
||||
user_id = await sessions_of(request).user_id_for(token)
|
||||
if user_id is None:
|
||||
raise Unauthorized("Сессия недействительна, истекла или отозвана")
|
||||
return user_id
|
||||
|
||||
|
||||
def path_uuid(request: Request, name: str) -> str:
|
||||
return validation.parse_uuid(request.path_params[name], name)
|
||||
|
||||
|
||||
async def read_json(request: Request) -> dict[str, Any]:
|
||||
content_type = request.headers.get("content-type", "")
|
||||
media_type, _, params = content_type.partition(";")
|
||||
if media_type.strip().lower() != "application/json":
|
||||
raise InvalidRequest("Content-Type должен быть application/json")
|
||||
charset = params.strip().lower()
|
||||
if charset and charset.replace(" ", "") not in {"charset=utf-8", "charset=\"utf-8\""}:
|
||||
raise InvalidRequest("Поддерживается только кодировка UTF-8")
|
||||
limit: int = request.app.state.settings.max_body_bytes
|
||||
declared = request.headers.get("content-length")
|
||||
if declared is not None and declared.isdigit() and int(declared) > limit:
|
||||
raise PayloadTooLarge()
|
||||
chunks: list[bytes] = []
|
||||
size = 0
|
||||
async for chunk in request.stream():
|
||||
size += len(chunk)
|
||||
if size > limit:
|
||||
raise PayloadTooLarge()
|
||||
chunks.append(chunk)
|
||||
return validation.parse_json_object(b"".join(chunks))
|
||||
|
||||
|
||||
def page_params(request: Request) -> tuple[int, str | None]:
|
||||
query = request.query_params.multi_items()
|
||||
limit = validation.parse_limit(validation.single_query_param(query, "limit"))
|
||||
raw_cursor = validation.single_query_param(query, "cursor")
|
||||
if raw_cursor is not None and raw_cursor == "":
|
||||
raise InvalidRequest("cursor не может быть пустым")
|
||||
return limit, raw_cursor
|
||||
|
||||
|
||||
async def reject_body(request: Request) -> None:
|
||||
"""POST /auth/sessions takes credentials only from the Basic header; any body is 400."""
|
||||
if request.headers.get("transfer-encoding") or request.headers.get("content-length", "0") != "0":
|
||||
raise InvalidRequest("Вход принимает только заголовок Authorization: Basic, без тела")
|
||||
async for chunk in request.stream():
|
||||
if chunk:
|
||||
raise InvalidRequest("Вход принимает только заголовок Authorization: Basic, без тела")
|
||||
|
||||
|
||||
def idempotency_key(request: Request) -> str:
|
||||
values = request.headers.getlist("idempotency-key")
|
||||
return validation.idempotency_key(values[0] if len(values) == 1 else None)
|
||||
|
||||
|
||||
def request_fingerprint(text: str, attachment_ids: list[str]) -> bytes:
|
||||
"""SHA-256 of the canonical body: key order and an absent attachment_ids vs [] do not matter."""
|
||||
canonical = json.dumps({"attachment_ids": attachment_ids, "text": text},
|
||||
ensure_ascii=False, sort_keys=True, separators=(",", ":"))
|
||||
return hashlib.sha256(canonical.encode("utf-8")).digest()
|
||||
|
||||
|
||||
def ok(body: Any, status: int = 200) -> JSONResponse:
|
||||
return JSONResponse(body, status_code=status)
|
||||
|
||||
|
||||
# --------------------------------------------------------------------------
|
||||
# Handlers
|
||||
# --------------------------------------------------------------------------
|
||||
async def live(request: Request) -> JSONResponse:
|
||||
return ok({"status": "ok"})
|
||||
|
||||
|
||||
async def ready(request: Request) -> JSONResponse:
|
||||
"""200 only if the mandatory synchronous path can be served: PostgreSQL and,
|
||||
with SESSION_STORE=valkey, Valkey answer. 503 while draining after SIGTERM."""
|
||||
if request.app.state.draining:
|
||||
raise Unavailable("Сервер завершает работу")
|
||||
db_ok, store_ok = await asyncio.gather(db_of(request).ping(), sessions_of(request).ping())
|
||||
if not db_ok:
|
||||
raise Unavailable("PostgreSQL недоступен")
|
||||
if not store_ok:
|
||||
raise Unavailable("Хранилище сессий недоступно")
|
||||
return ok({"status": "ok"})
|
||||
|
||||
|
||||
async def register_user(request: Request) -> JSONResponse:
|
||||
data = await read_json(request)
|
||||
validation.check_fields(data, {"username", "display_name", "password"})
|
||||
username = validation.username_field(data)
|
||||
display_name = validation.text_field(data, "display_name", 100)
|
||||
password = validation.password_field(data)
|
||||
password_hash = await passwords.hash_password(password)
|
||||
user = await repo.create_user(db_of(request), username, display_name, password_hash)
|
||||
log.info("user_created", extra={"fields": {"request_id": request.state.request_id, "user_id": user.id}})
|
||||
return ok(user.to_json(), 201)
|
||||
|
||||
|
||||
async def create_session(request: Request) -> JSONResponse:
|
||||
"""Login: only `Authorization: Basic`; wrong user and wrong password give the same 401."""
|
||||
username, password = authhdr.parse_basic(request.headers.get("authorization"))
|
||||
await reject_body(request)
|
||||
found = await repo.get_credentials(db_of(request), username)
|
||||
user, stored_hash = found if found else (None, None)
|
||||
if not await passwords.verify_password(stored_hash, password) or user is None:
|
||||
log.info("login_failed", extra={"fields": {"request_id": request.state.request_id}})
|
||||
raise Unauthorized("Неверное имя пользователя или пароль", challenge=BASIC_CHALLENGE)
|
||||
ttl = request.app.state.settings.session_ttl_seconds
|
||||
token, expires_at = await sessions_of(request).create(user.id, ttl)
|
||||
log.info("session_created", extra={"fields": {"request_id": request.state.request_id, "user_id": user.id}})
|
||||
return ok({"token": token, "token_type": "Bearer", "expires_at": rfc3339(expires_at), "user": user.to_json()}, 201)
|
||||
|
||||
|
||||
async def revoke_session(request: Request) -> Response:
|
||||
token = authhdr.parse_bearer(request.headers.get("authorization"))
|
||||
if not await sessions_of(request).revoke(token):
|
||||
raise Unauthorized("Сессия недействительна, истекла или отозвана")
|
||||
log.info("session_revoked", extra={"fields": {"request_id": request.state.request_id}})
|
||||
return Response(status_code=204)
|
||||
|
||||
|
||||
async def me(request: Request) -> JSONResponse:
|
||||
user_id = await authenticate(request)
|
||||
user = await repo.get_user(db_of(request), user_id)
|
||||
if user is None:
|
||||
raise Unauthorized("Сессия недействительна, истекла или отозвана")
|
||||
return ok(user.to_json())
|
||||
|
||||
|
||||
async def create_chat(request: Request) -> JSONResponse:
|
||||
user_id = await authenticate(request)
|
||||
data = await read_json(request)
|
||||
validation.check_fields(data, {"title", "member_ids"})
|
||||
title = validation.text_field(data, "title", 100)
|
||||
member_ids = validation.member_ids_field(data)
|
||||
chat = await repo.create_chat(db_of(request), user_id, title, member_ids)
|
||||
log.info("chat_created", extra={"fields": {"request_id": request.state.request_id, "chat_id": chat.id, "members": len(member_ids)}})
|
||||
return ok(chat.to_json(), 201)
|
||||
|
||||
|
||||
async def list_chats(request: Request) -> JSONResponse:
|
||||
user_id = await authenticate(request)
|
||||
limit, raw_cursor = page_params(request)
|
||||
after = cursor.decode(raw_cursor, cursor.CHATS, user_id) if raw_cursor else None
|
||||
items, has_more = await repo.list_chats(db_of(request), user_id, after, limit)
|
||||
next_cursor = cursor.encode(cursor.CHATS, user_id, items[-1].key) if has_more else None
|
||||
return ok({"items": [c.to_json() for c in items], "next_cursor": next_cursor})
|
||||
|
||||
|
||||
async def get_chat(request: Request) -> JSONResponse:
|
||||
user_id = await authenticate(request)
|
||||
chat_id = path_uuid(request, "chat_id")
|
||||
return ok((await repo.get_chat(db_of(request), user_id, chat_id)).to_json())
|
||||
|
||||
|
||||
def parse_message(data: dict[str, Any]) -> tuple[str, list[str]]:
|
||||
validation.check_fields(data, {"text"}, frozenset({"attachment_ids"}))
|
||||
text = validation.text_field(data, "text", 4000)
|
||||
attachment_ids = validation.attachment_ids_field(data)
|
||||
if attachment_ids:
|
||||
# Lab 3 has no attachments: every referenced ID is unknown (lab 4 semantics: 404).
|
||||
raise NotFound("Вложение не найдено")
|
||||
return text, attachment_ids
|
||||
|
||||
|
||||
async def create_message(request: Request) -> JSONResponse:
|
||||
"""Order of answers for both write paths: 401 -> 400 bad chat id -> 404 outsider -> 400 key/body
|
||||
-> 409 other body with the same key -> 201 (new or replayed). sender_id comes from the session."""
|
||||
user_id = await authenticate(request)
|
||||
chat_id = path_uuid(request, "chat_id")
|
||||
db = db_of(request)
|
||||
settings = request.app.state.settings
|
||||
if settings.message_write_path == "multi":
|
||||
await repo.ensure_member(db, user_id, chat_id) # outsiders get 404 before key/body are looked at
|
||||
key = idempotency_key(request)
|
||||
text, attachment_ids = parse_message(await read_json(request))
|
||||
result = await repo.post_message(db, user_id, chat_id, key, request_fingerprint(text, attachment_ids), text,
|
||||
settings.idempotency_ttl_seconds)
|
||||
else:
|
||||
try:
|
||||
key = idempotency_key(request)
|
||||
text, attachment_ids = parse_message(await read_json(request))
|
||||
except ApiError:
|
||||
# Rare error path only: an outsider must still get 404, not a hint about the body.
|
||||
await repo.ensure_member(db, user_id, chat_id)
|
||||
raise
|
||||
# Membership check, key claim, chat clock and message: one SQL call, one transaction.
|
||||
result = await repo.post_message_single_call(db, user_id, chat_id, key,
|
||||
request_fingerprint(text, attachment_ids), text,
|
||||
settings.idempotency_ttl_seconds)
|
||||
if result.replayed:
|
||||
log.info("idempotent_replay", extra={"fields": {"request_id": request.state.request_id,
|
||||
"message_id": result.message.id}})
|
||||
return ok(result.message.to_json(), 201)
|
||||
|
||||
|
||||
async def list_messages(request: Request) -> JSONResponse:
|
||||
user_id = await authenticate(request)
|
||||
chat_id = path_uuid(request, "chat_id")
|
||||
db = db_of(request)
|
||||
await repo.ensure_member(db, user_id, chat_id) # 404 for outsiders before cursor details leak
|
||||
limit, raw_cursor = page_params(request)
|
||||
after = cursor.decode(raw_cursor, cursor.MESSAGES, chat_id) if raw_cursor else None
|
||||
items, has_more = await repo.list_messages(db, chat_id, after, limit)
|
||||
next_cursor = cursor.encode(cursor.MESSAGES, chat_id, items[-1].key) if has_more else None
|
||||
return ok({"items": [m.to_json() for m in items], "next_cursor": next_cursor})
|
||||
|
||||
|
||||
def create_app(settings: Settings | None = None) -> Starlette:
|
||||
settings = settings or Settings.from_env()
|
||||
db = Database(settings)
|
||||
sessions = build_store(settings, db)
|
||||
|
||||
async def purge_loop() -> None:
|
||||
"""Housekeeping: expired Idempotency-Key rows (older than IDEMPOTENCY_TTL_SECONDS)."""
|
||||
while True:
|
||||
await asyncio.sleep(600)
|
||||
try:
|
||||
deleted = await repo.purge_idempotency_keys(db, settings.idempotency_ttl_seconds)
|
||||
if deleted:
|
||||
log.info("idempotency_keys_purged", extra={"fields": {"deleted": deleted}})
|
||||
except ApiError:
|
||||
pass # database unavailable right now; next round
|
||||
|
||||
stats_log = logging.getLogger("messenger.stats")
|
||||
|
||||
async def stats_loop(every: float) -> None:
|
||||
"""Experiments only: DB pool saturation (waiting requests, wait time) and process CPU."""
|
||||
cpu = time.process_time()
|
||||
db.pool.pop_stats()
|
||||
while True:
|
||||
await asyncio.sleep(every)
|
||||
stats = db.pool.pop_stats()
|
||||
now_cpu = time.process_time()
|
||||
stats_log.info("pool_stats", extra={"fields": {
|
||||
"interval_s": every, "cpu_s": round(now_cpu - cpu, 4),
|
||||
**{k: stats.get(k, 0) for k in ("pool_size", "pool_available", "requests_waiting", "requests_num",
|
||||
"requests_queued", "requests_wait_ms", "usage_ms", "requests_errors")}}})
|
||||
cpu = now_cpu
|
||||
|
||||
@contextlib.asynccontextmanager
|
||||
async def lifespan(app: Starlette):
|
||||
passwords.configure(settings.hash_concurrency)
|
||||
await db.open()
|
||||
log.info("dependencies_configured", extra={"fields": {
|
||||
"session_store": sessions.name, "session_ttl_s": settings.session_ttl_seconds,
|
||||
"db_pool_max": settings.db_pool_max, "db_timeout_s": settings.db_timeout_seconds,
|
||||
"instance_id": settings.instance_id, "drain_s": settings.drain_seconds,
|
||||
"message_write_path": settings.message_write_path}})
|
||||
tasks = [asyncio.create_task(purge_loop())]
|
||||
if settings.stats_log_seconds > 0:
|
||||
tasks.append(asyncio.create_task(stats_loop(settings.stats_log_seconds)))
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
for task in tasks:
|
||||
task.cancel()
|
||||
await sessions.close()
|
||||
await db.close()
|
||||
|
||||
routes = [
|
||||
Route("/health/live", live, methods=["GET"]),
|
||||
Route("/health/ready", ready, methods=["GET"]),
|
||||
Route("/api/v1/users", register_user, methods=["POST"]),
|
||||
Route("/api/v1/users/me", me, methods=["GET"]),
|
||||
Route("/api/v1/auth/sessions", create_session, methods=["POST"]),
|
||||
Route("/api/v1/auth/sessions/current", revoke_session, methods=["DELETE"]),
|
||||
Route("/api/v1/chats", create_chat, methods=["POST"]),
|
||||
Route("/api/v1/chats", list_chats, methods=["GET"]),
|
||||
Route("/api/v1/chats/{chat_id}", get_chat, methods=["GET"]),
|
||||
Route("/api/v1/chats/{chat_id}/messages", create_message, methods=["POST"]),
|
||||
Route("/api/v1/chats/{chat_id}/messages", list_messages, methods=["GET"]),
|
||||
]
|
||||
if settings.docs_enabled: # Swagger UI for manual experiments, not part of the contract
|
||||
routes += [
|
||||
Route("/docs", docs.swagger_page, methods=["GET"]),
|
||||
Route("/openapi.json", docs.openapi_json, methods=["GET"]),
|
||||
]
|
||||
app = Starlette(
|
||||
routes=routes,
|
||||
exception_handlers={ApiError: api_error_handler, HTTPException: http_exception_handler},
|
||||
lifespan=lifespan,
|
||||
)
|
||||
app.add_middleware(RequestContextMiddleware, instance_id=settings.instance_id)
|
||||
app.state.settings = settings
|
||||
app.state.db = db
|
||||
app.state.sessions = sessions
|
||||
app.state.draining = False
|
||||
return app
|
||||
@@ -0,0 +1,48 @@
|
||||
"""Parsing of the Authorization header. Any malformed value is 401, never 500.
|
||||
|
||||
Basic (RFC 7617) is accepted only by POST /auth/sessions; everything else needs
|
||||
Bearer. Basic is an encoding, not encryption: plain HTTP is acceptable only for the
|
||||
local synthetic lab stand (lab 7 adds TLS).
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import binascii
|
||||
import re
|
||||
|
||||
from .errors import BASIC_CHALLENGE, BEARER_CHALLENGE, Unauthorized
|
||||
from .validation import USERNAME_RE
|
||||
|
||||
B64_RE = re.compile(r"[A-Za-z0-9+/]+={0,2}")
|
||||
TOKEN_RE = re.compile(r"[A-Za-z0-9_-]{32,128}") # token_urlsafe(32) gives 43 characters
|
||||
|
||||
|
||||
def _split(header: str | None) -> tuple[str, str]:
|
||||
if not header:
|
||||
return "", ""
|
||||
scheme, _, value = header.strip().partition(" ")
|
||||
return scheme.lower(), value.strip()
|
||||
|
||||
|
||||
def parse_basic(header: str | None) -> tuple[str, str]:
|
||||
"""(username, password) from `Basic base64(username:password)`; the password may contain ':'."""
|
||||
invalid = Unauthorized("Неверное имя пользователя или пароль", challenge=BASIC_CHALLENGE)
|
||||
scheme, value = _split(header)
|
||||
if scheme != "basic" or B64_RE.fullmatch(value) is None or len(value) > 1024:
|
||||
raise invalid
|
||||
try:
|
||||
decoded = base64.b64decode(value, validate=True).decode("utf-8")
|
||||
except (binascii.Error, ValueError, UnicodeDecodeError):
|
||||
raise invalid from None
|
||||
username, sep, password = decoded.partition(":") # split at the FIRST colon
|
||||
if not sep or USERNAME_RE.fullmatch(username) is None or not password:
|
||||
raise invalid
|
||||
return username, password
|
||||
|
||||
|
||||
def parse_bearer(header: str | None) -> str:
|
||||
scheme, value = _split(header)
|
||||
if scheme != "bearer" or TOKEN_RE.fullmatch(value) is None:
|
||||
# Missing header, Basic on a business endpoint, garbage token: one answer.
|
||||
raise Unauthorized("Требуется действительная Bearer-сессия", challenge=BEARER_CHALLENGE)
|
||||
return value
|
||||
@@ -0,0 +1,168 @@
|
||||
"""Runtime configuration read from environment variables (12-factor style).
|
||||
|
||||
Non-secret settings have safe defaults. Connection strings with passwords are
|
||||
read from files (`DATABASE_URL_FILE`, `CACHE_URL_FILE` → Docker secrets in
|
||||
/run/secrets), so they never appear in `docker inspect`, compose files or logs.
|
||||
Invalid values stop the process at startup with a clear message.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
import re
|
||||
import socket
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
class ConfigError(ValueError):
|
||||
pass
|
||||
|
||||
|
||||
INSTANCE_ID_RE = re.compile(r"[A-Za-z0-9._-]{1,64}")
|
||||
|
||||
|
||||
def _instance_id() -> str:
|
||||
"""X-Instance-Id of this process: INSTANCE_ID or the container hostname.
|
||||
|
||||
Docker sets the hostname of a container to its short container ID, so the value
|
||||
is unique per replica, stable for the life of the process and maps 1:1 to a real
|
||||
container (`docker inspect`), which scripts/stop-one.sh relies on.
|
||||
"""
|
||||
value = os.environ.get("INSTANCE_ID", "").strip() or socket.gethostname()
|
||||
if INSTANCE_ID_RE.fullmatch(value) is None:
|
||||
raise ConfigError("INSTANCE_ID must match [A-Za-z0-9._-]{1,64}")
|
||||
return value
|
||||
|
||||
|
||||
def _int(name: str, default: int, low: int, high: int) -> int:
|
||||
raw = os.environ.get(name)
|
||||
if raw is None or raw.strip() == "":
|
||||
return default
|
||||
try:
|
||||
value = int(raw.strip())
|
||||
except ValueError:
|
||||
raise ConfigError(f"{name} must be an integer, got {raw!r}") from None
|
||||
if not low <= value <= high:
|
||||
raise ConfigError(f"{name} must be in [{low}, {high}], got {value}")
|
||||
return value
|
||||
|
||||
|
||||
def _float(name: str, default: float, low: float, high: float) -> float:
|
||||
raw = os.environ.get(name)
|
||||
if raw is None or raw.strip() == "":
|
||||
return default
|
||||
try:
|
||||
value = float(raw.strip())
|
||||
except ValueError:
|
||||
raise ConfigError(f"{name} must be a number, got {raw!r}") from None
|
||||
if not low <= value <= high:
|
||||
raise ConfigError(f"{name} must be in [{low}, {high}], got {value}")
|
||||
return value
|
||||
|
||||
|
||||
def _bool(name: str, default: bool) -> bool:
|
||||
raw = os.environ.get(name, "").strip().lower()
|
||||
if raw == "":
|
||||
return default
|
||||
if raw not in {"true", "false", "1", "0"}:
|
||||
raise ConfigError(f"{name} must be true/false, got {raw!r}")
|
||||
return raw in {"true", "1"}
|
||||
|
||||
|
||||
def _choice(name: str, default: str, allowed: set[str]) -> str:
|
||||
value = os.environ.get(name, "").strip().lower() or default
|
||||
if value not in allowed:
|
||||
raise ConfigError(f"{name} must be one of {sorted(allowed)}, got {value!r}")
|
||||
return value
|
||||
|
||||
|
||||
def _secret(name: str, required: bool) -> str:
|
||||
"""Value of <NAME>_FILE (preferred, Docker secret) or <NAME> (local runs/tests)."""
|
||||
path = os.environ.get(f"{name}_FILE", "").strip()
|
||||
if path:
|
||||
try:
|
||||
value = Path(path).read_text(encoding="utf-8").strip()
|
||||
except OSError as exc:
|
||||
# The message names the file, never its content.
|
||||
raise ConfigError(f"{name}_FILE={path} cannot be read: {exc.strerror}") from None
|
||||
else:
|
||||
value = os.environ.get(name, "").strip()
|
||||
if required and not value:
|
||||
raise ConfigError(f"{name}_FILE (or {name}) must be set")
|
||||
return value
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Settings:
|
||||
host: str
|
||||
port: int
|
||||
env: str
|
||||
log_level: str
|
||||
shutdown_timeout: float
|
||||
max_body_bytes: int
|
||||
docs_enabled: bool = True
|
||||
database_url: str = ""
|
||||
session_store: str = "postgres"
|
||||
cache_url: str = ""
|
||||
session_ttl_seconds: int = 3600
|
||||
db_pool_min: int = 1
|
||||
db_pool_max: int = 10
|
||||
db_timeout_seconds: float = 3.0
|
||||
cache_timeout_seconds: float = 0.5
|
||||
hash_concurrency: int = 2
|
||||
instance_id: str = "local"
|
||||
drain_seconds: float = 0.0
|
||||
idempotency_ttl_seconds: int = 86400
|
||||
stats_log_seconds: float = 0.0
|
||||
message_write_path: str = "function"
|
||||
|
||||
@property
|
||||
def db_statement_timeout_ms(self) -> int:
|
||||
# The server stops a statement a bit before the client-side deadline fires.
|
||||
return max(100, int(self.db_timeout_seconds * 1000) - 500)
|
||||
|
||||
@classmethod
|
||||
def from_env(cls) -> "Settings":
|
||||
log_level = os.environ.get("LOG_LEVEL", "info").strip().lower() or "info"
|
||||
if log_level not in {"debug", "info", "warning", "error"}:
|
||||
raise ConfigError(f"LOG_LEVEL must be debug/info/warning/error, got {log_level!r}")
|
||||
session_store = os.environ.get("SESSION_STORE", "postgres").strip().lower() or "postgres"
|
||||
if session_store == "redis":
|
||||
session_store = "valkey" # same protocol; the course allows Redis OR Valkey
|
||||
if session_store not in {"postgres", "valkey"}:
|
||||
raise ConfigError(f"SESSION_STORE must be postgres or valkey, got {session_store!r}")
|
||||
db_pool_min = _int("DB_POOL_MIN", 1, 0, 50)
|
||||
db_pool_max = _int("DB_POOL_MAX", 10, 1, 100)
|
||||
if db_pool_min > db_pool_max:
|
||||
raise ConfigError("DB_POOL_MIN must not exceed DB_POOL_MAX")
|
||||
return cls(
|
||||
host=os.environ.get("APP_HOST", "0.0.0.0").strip() or "0.0.0.0",
|
||||
port=_int("APP_PORT", 8080, 1, 65535),
|
||||
env=os.environ.get("APP_ENV", "lab").strip() or "lab",
|
||||
log_level=log_level,
|
||||
# DRAIN_SECONDS + SHUTDOWN_TIMEOUT_SECONDS must stay below compose stop_grace_period.
|
||||
shutdown_timeout=_float("SHUTDOWN_TIMEOUT_SECONDS", 8.0, 0.0, 30.0),
|
||||
max_body_bytes=_int("MAX_BODY_BYTES", 64 * 1024, 1024, 1024 * 1024),
|
||||
docs_enabled=_bool("DOCS_ENABLED", True),
|
||||
database_url=_secret("DATABASE_URL", required=True),
|
||||
session_store=session_store,
|
||||
cache_url=_secret("CACHE_URL", required=session_store == "valkey"),
|
||||
# Contract name. Minimum 1 s so the expiry test can use 2 s.
|
||||
session_ttl_seconds=_int("SESSION_TTL_SECONDS", 3600, 1, 30 * 24 * 3600),
|
||||
db_pool_min=db_pool_min,
|
||||
db_pool_max=db_pool_max,
|
||||
db_timeout_seconds=_float("DB_TIMEOUT_SECONDS", 3.0, 0.5, 30.0),
|
||||
cache_timeout_seconds=_float("CACHE_TIMEOUT_SECONDS", 0.5, 0.05, 10.0),
|
||||
hash_concurrency=_int("PASSWORD_HASH_CONCURRENCY", 2, 1, 16),
|
||||
instance_id=_instance_id(),
|
||||
# After SIGTERM: readiness is 503 but requests are still served for this long,
|
||||
# so the load balancer stops routing here before the listener closes.
|
||||
drain_seconds=_float("DRAIN_SECONDS", 0.0, 0.0, 10.0),
|
||||
# Contract: a replay within 24 h returns the original result.
|
||||
idempotency_ttl_seconds=_int("IDEMPOTENCY_TTL_SECONDS", 86400, 86400, 30 * 24 * 3600),
|
||||
# Experiments (bench/track1): log DB pool counters and process CPU every N s; 0 = off.
|
||||
stats_log_seconds=_float("STATS_LOG_SECONDS", 0.0, 0.0, 60.0),
|
||||
# Track 1: `function` = one SQL call per send (migration 0003), `multi` = the previous
|
||||
# multi-statement transaction; both have identical semantics (A/B on one image).
|
||||
message_write_path=_choice("MESSAGE_WRITE_PATH", "function", {"function", "multi"}),
|
||||
)
|
||||
@@ -0,0 +1,56 @@
|
||||
"""Opaque keyset cursors.
|
||||
|
||||
A cursor encodes the sort key `(created_at, id)` of the last item on the page
|
||||
plus the resource it belongs to, e.g. `v1|m|<chat_id>|<ts_us>|<message_id>`,
|
||||
packed in URL-safe base64. The client treats it as an opaque string; the server
|
||||
rejects cursors of another chat/user (400) and any non-canonical encoding.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import re
|
||||
|
||||
from .errors import InvalidRequest
|
||||
from .validation import UUID_RE
|
||||
|
||||
B64_RE = re.compile(r"[A-Za-z0-9_-]{1,512}")
|
||||
TS_RE = re.compile(r"[0-9]{1,19}")
|
||||
|
||||
# Latest instant both Python datetime and PostgreSQL accept: 9999-12-31T23:59:59.999999Z.
|
||||
MAX_TS_US = 253402300799999999
|
||||
|
||||
MESSAGES = "m"
|
||||
CHATS = "c"
|
||||
|
||||
SortKey = tuple[int, str]
|
||||
|
||||
|
||||
def encode(scope: str, owner_id: str, key: SortKey) -> str:
|
||||
ts, item_id = key
|
||||
raw = f"v1|{scope}|{owner_id}|{ts}|{item_id}".encode("ascii")
|
||||
return base64.urlsafe_b64encode(raw).rstrip(b"=").decode("ascii")
|
||||
|
||||
|
||||
def decode(cursor: str, scope: str, owner_id: str) -> SortKey:
|
||||
"""Return the sort key stored in the cursor or raise InvalidRequest."""
|
||||
bad = InvalidRequest("Неверный cursor")
|
||||
if B64_RE.fullmatch(cursor) is None:
|
||||
raise bad
|
||||
try:
|
||||
raw = base64.urlsafe_b64decode(cursor + "=" * (-len(cursor) % 4)).decode("ascii")
|
||||
except (ValueError, UnicodeDecodeError):
|
||||
raise bad from None
|
||||
parts = raw.split("|")
|
||||
if len(parts) != 5 or parts[0] != "v1":
|
||||
raise bad
|
||||
_, got_scope, got_owner, ts, item_id = parts
|
||||
if TS_RE.fullmatch(ts) is None or UUID_RE.fullmatch(item_id) is None or item_id != item_id.lower():
|
||||
raise bad
|
||||
key = (int(ts), item_id)
|
||||
if key[0] > MAX_TS_US:
|
||||
raise bad # out-of-range time must be 400, not an overflow (500) in datetime/SQL
|
||||
if encode(got_scope, got_owner, key) != cursor:
|
||||
raise bad # only the canonical form produced by the server is accepted
|
||||
if got_scope != scope or got_owner != owner_id:
|
||||
raise InvalidRequest("cursor относится к другому ресурсу")
|
||||
return key
|
||||
@@ -0,0 +1,131 @@
|
||||
"""PostgreSQL access: a bounded connection pool and a hard deadline for every unit of work.
|
||||
|
||||
Three independent limits keep a request from hanging when the database misbehaves:
|
||||
|
||||
* pool `timeout` — waiting for a free connection (pool exhausted);
|
||||
* server `statement_timeout` / `lock_timeout` — PostgreSQL itself cancels a slow
|
||||
statement or a statement stuck behind a lock;
|
||||
* client deadline — covers what the server cannot report, e.g. a
|
||||
paused/frozen PostgreSQL container or a dead network.
|
||||
|
||||
The client deadline needs care. Every unit of work runs as its own task; at the deadline
|
||||
the REQUEST stops waiting and gets 503 at once, but the task is NOT cancelled: cancelling a
|
||||
psycopg query makes psycopg send a cancel request, and libpq starts that connection
|
||||
(PQcancelStart) synchronously in the event-loop thread, resolving the host name first.
|
||||
When the API has lost its network to PostgreSQL (lab 3 review scenario: `docker network
|
||||
disconnect`), that blocked the WHOLE replica, /health/live included, for ~1 minute.
|
||||
The abandoned task instead ends by itself, bounded by:
|
||||
|
||||
* TCP keepalives + `tcp_user_timeout` (= the deadline): a dead peer or a vanished network
|
||||
breaks the socket within seconds, the connection is discarded by the pool;
|
||||
* `statement_timeout` / `lock_timeout` on a live but slow server;
|
||||
* a frozen (paused) server keeps the connection until it resumes; meanwhile new work waits
|
||||
at most the pool timeout and gets 503.
|
||||
|
||||
Any failure becomes 503 `unavailable` — never 401, never a hang.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import logging
|
||||
from collections.abc import Awaitable, Callable
|
||||
from typing import TypeVar
|
||||
|
||||
import psycopg
|
||||
from psycopg_pool import AsyncConnectionPool, PoolTimeout
|
||||
|
||||
from .config import Settings
|
||||
from .errors import Unavailable
|
||||
|
||||
log = logging.getLogger("messenger.db")
|
||||
|
||||
T = TypeVar("T")
|
||||
Work = Callable[[psycopg.AsyncConnection], Awaitable[T]]
|
||||
|
||||
# Errors that mean "the database is unavailable right now", not "the request is wrong".
|
||||
# QueryCanceled (statement_timeout), LockNotAvailable (lock_timeout) and AdminShutdown
|
||||
# (server restart) are subclasses of OperationalError.
|
||||
UNAVAILABLE_ERRORS = (psycopg.OperationalError, psycopg.InterfaceError, PoolTimeout, TimeoutError)
|
||||
|
||||
|
||||
def _forget(task: asyncio.Task) -> None:
|
||||
"""Done-callback for abandoned tasks: retrieve the result so nothing is logged as 'never retrieved'."""
|
||||
if not task.cancelled():
|
||||
task.exception()
|
||||
|
||||
|
||||
class Database:
|
||||
def __init__(self, settings: Settings):
|
||||
self.deadline = settings.db_timeout_seconds
|
||||
options = (
|
||||
f"-c statement_timeout={settings.db_statement_timeout_ms} "
|
||||
f"-c lock_timeout={max(100, settings.db_statement_timeout_ms // 2)} "
|
||||
"-c idle_in_transaction_session_timeout=10000"
|
||||
)
|
||||
self.pool = AsyncConnectionPool(
|
||||
conninfo=settings.database_url,
|
||||
min_size=settings.db_pool_min,
|
||||
max_size=settings.db_pool_max,
|
||||
timeout=settings.db_timeout_seconds, # max wait for a free connection
|
||||
open=False,
|
||||
name="messenger",
|
||||
check=AsyncConnectionPool.check_connection, # drop connections broken by a DB restart
|
||||
kwargs={
|
||||
"options": options,
|
||||
"connect_timeout": max(1, int(settings.db_timeout_seconds)),
|
||||
"application_name": "messenger-api",
|
||||
# Autocommit connections: a single statement (post_message function, ping) is its
|
||||
# own transaction with ONE round trip; multi-statement work uses an explicit
|
||||
# conn.transaction() (BEGIN ... COMMIT), see run().
|
||||
"autocommit": True,
|
||||
# Detect a dead peer / lost network without any cancel request (see module doc).
|
||||
"keepalives": 1,
|
||||
"keepalives_idle": 5,
|
||||
"keepalives_interval": 1,
|
||||
"keepalives_count": 3,
|
||||
"tcp_user_timeout": int(settings.db_timeout_seconds * 1000),
|
||||
},
|
||||
)
|
||||
|
||||
async def open(self) -> None:
|
||||
# wait=False: the API starts (liveness 200) even if PostgreSQL is not up yet;
|
||||
# readiness stays 503 until the database answers.
|
||||
await self.pool.open(wait=False)
|
||||
|
||||
async def close(self) -> None:
|
||||
await self.pool.close(timeout=2)
|
||||
|
||||
async def run(self, work: Work[T], deadline: float | None = None, *, transaction: bool = True) -> T:
|
||||
"""Run `work(conn)` in one transaction (COMMIT on success, ROLLBACK on any exception).
|
||||
transaction=False: `work` issues exactly ONE statement, which autocommit makes atomic.
|
||||
|
||||
Domain errors raised by `work` (NotFound, Conflict, ...) propagate unchanged;
|
||||
database unavailability and the deadline become Unavailable (503).
|
||||
"""
|
||||
async def unit() -> T:
|
||||
async with self.pool.connection() as conn:
|
||||
if not transaction:
|
||||
return await work(conn)
|
||||
async with conn.transaction():
|
||||
return await work(conn)
|
||||
|
||||
task = asyncio.ensure_future(unit())
|
||||
try:
|
||||
return await asyncio.wait_for(asyncio.shield(task), deadline or self.deadline)
|
||||
except UNAVAILABLE_ERRORS as exc:
|
||||
log.warning("database_unavailable", extra={"fields": {"error": type(exc).__name__}})
|
||||
raise Unavailable("База данных временно недоступна") from None
|
||||
finally:
|
||||
if not task.done():
|
||||
# Deadline or cancelled request: stop waiting now. The task is left to finish
|
||||
# on its own (no task.cancel(): that would trigger psycopg's blocking cancel).
|
||||
task.add_done_callback(_forget)
|
||||
|
||||
async def ping(self, timeout: float = 1.0) -> bool:
|
||||
async def select_one(conn: psycopg.AsyncConnection) -> None:
|
||||
await conn.execute("SELECT 1")
|
||||
try:
|
||||
await self.run(select_one, deadline=timeout, transaction=False)
|
||||
return True
|
||||
except Unavailable:
|
||||
return False
|
||||
@@ -0,0 +1,57 @@
|
||||
"""Swagger UI for manual experiments: http://127.0.0.1:8080/docs
|
||||
|
||||
The page shows the course contract contracts/openapi.json as is (it is not
|
||||
generated from our code). Only `servers` is replaced with "/" so that the
|
||||
"Try it out" requests go to the same host the page was opened from.
|
||||
Swagger UI itself (JS/CSS) is loaded by the browser from the jsDelivr CDN.
|
||||
Disable with DOCS_ENABLED=false.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
from functools import lru_cache
|
||||
from pathlib import Path
|
||||
|
||||
from starlette.requests import Request
|
||||
from starlette.responses import HTMLResponse, JSONResponse
|
||||
|
||||
SPEC_PATH = Path(__file__).resolve().parents[2] / "contracts" / "openapi.json"
|
||||
SWAGGER_UI = "https://cdn.jsdelivr.net/npm/swagger-ui-dist@5.33.0"
|
||||
|
||||
PAGE = f"""<!doctype html>
|
||||
<html lang="ru">
|
||||
<head>
|
||||
<meta charset="utf-8">
|
||||
<title>Messenger lab1 — Swagger UI</title>
|
||||
<link rel="stylesheet" href="{SWAGGER_UI}/swagger-ui.css">
|
||||
</head>
|
||||
<body>
|
||||
<div id="swagger-ui"></div>
|
||||
<script src="{SWAGGER_UI}/swagger-ui-bundle.js"></script>
|
||||
<script>
|
||||
window.ui = SwaggerUIBundle({{
|
||||
url: "/openapi.json",
|
||||
dom_id: "#swagger-ui",
|
||||
tryItOutEnabled: true, // fields are editable right away
|
||||
persistAuthorization: true, // X-User-Id from "Authorize" survives page reload
|
||||
displayRequestDuration: true,
|
||||
}});
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
"""
|
||||
|
||||
|
||||
@lru_cache(maxsize=1)
|
||||
def _spec() -> dict:
|
||||
spec = json.loads(SPEC_PATH.read_text(encoding="utf-8"))
|
||||
spec["servers"] = [{"url": "/", "description": "этот сервер"}]
|
||||
return spec
|
||||
|
||||
|
||||
async def swagger_page(request: Request) -> HTMLResponse:
|
||||
return HTMLResponse(PAGE)
|
||||
|
||||
|
||||
async def openapi_json(request: Request) -> JSONResponse:
|
||||
return JSONResponse(_spec())
|
||||
@@ -0,0 +1,65 @@
|
||||
"""Domain errors mapped 1:1 to the contract error codes (contracts/README.md)."""
|
||||
from __future__ import annotations
|
||||
|
||||
BEARER_CHALLENGE = 'Bearer realm="messenger"'
|
||||
BASIC_CHALLENGE = 'Basic realm="messenger", charset="UTF-8"'
|
||||
|
||||
|
||||
class ApiError(Exception):
|
||||
status = 500
|
||||
code = "internal_error"
|
||||
default_message = "Внутренняя ошибка сервера"
|
||||
|
||||
def __init__(self, message: str | None = None):
|
||||
self.message = message or self.default_message
|
||||
super().__init__(self.message)
|
||||
|
||||
@property
|
||||
def headers(self) -> dict[str, str]:
|
||||
return {}
|
||||
|
||||
|
||||
class InvalidRequest(ApiError):
|
||||
status = 400
|
||||
code = "invalid_request"
|
||||
default_message = "Неверный запрос"
|
||||
|
||||
|
||||
class Unauthorized(ApiError):
|
||||
"""401 with the challenge of the scheme the endpoint expects (RFC 7235)."""
|
||||
|
||||
status = 401
|
||||
code = "unauthorized"
|
||||
default_message = "Требуется действительная сессия"
|
||||
|
||||
def __init__(self, message: str | None = None, challenge: str = BEARER_CHALLENGE):
|
||||
super().__init__(message)
|
||||
self.challenge = challenge
|
||||
|
||||
@property
|
||||
def headers(self) -> dict[str, str]:
|
||||
return {"WWW-Authenticate": self.challenge}
|
||||
|
||||
|
||||
class NotFound(ApiError):
|
||||
status = 404
|
||||
code = "not_found"
|
||||
default_message = "Объект не найден"
|
||||
|
||||
|
||||
class Conflict(ApiError):
|
||||
status = 409
|
||||
code = "conflict"
|
||||
default_message = "Конфликт"
|
||||
|
||||
|
||||
class PayloadTooLarge(ApiError):
|
||||
status = 413
|
||||
code = "payload_too_large"
|
||||
default_message = "Тело запроса слишком большое"
|
||||
|
||||
|
||||
class Unavailable(ApiError):
|
||||
status = 503
|
||||
code = "unavailable"
|
||||
default_message = "Сервис временно недоступен"
|
||||
@@ -0,0 +1,19 @@
|
||||
"""Docker HEALTHCHECK probe without curl/wget in the image: `python -m messenger.healthcheck`."""
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
import sys
|
||||
import urllib.request
|
||||
|
||||
|
||||
def main() -> int:
|
||||
port = os.environ.get("APP_PORT", "8080").strip() or "8080"
|
||||
try:
|
||||
with urllib.request.urlopen(f"http://127.0.0.1:{port}/health/ready", timeout=2) as response:
|
||||
return 0 if response.status == 200 else 1
|
||||
except Exception:
|
||||
return 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
@@ -0,0 +1,36 @@
|
||||
"""JSON-lines logging to stdout. Logs never contain message texts or request bodies."""
|
||||
from __future__ import annotations
|
||||
|
||||
import datetime as dt
|
||||
import json
|
||||
import logging
|
||||
import sys
|
||||
|
||||
|
||||
class JsonFormatter(logging.Formatter):
|
||||
def format(self, record: logging.LogRecord) -> str:
|
||||
entry = {
|
||||
"ts": dt.datetime.fromtimestamp(record.created, dt.timezone.utc).isoformat(timespec="milliseconds").replace("+00:00", "Z"),
|
||||
"level": record.levelname.lower(),
|
||||
"logger": record.name,
|
||||
"event": record.getMessage(),
|
||||
}
|
||||
fields = getattr(record, "fields", None)
|
||||
if fields:
|
||||
entry.update(fields)
|
||||
if record.exc_info:
|
||||
# Stack traces go only to the server log, never to the HTTP client.
|
||||
entry["exception"] = self.formatException(record.exc_info)
|
||||
return json.dumps(entry, ensure_ascii=False)
|
||||
|
||||
|
||||
def configure(level: str) -> None:
|
||||
handler = logging.StreamHandler(sys.stdout)
|
||||
handler.setFormatter(JsonFormatter())
|
||||
root = logging.getLogger()
|
||||
root.handlers[:] = [handler]
|
||||
root.setLevel(level.upper())
|
||||
# uvicorn's own access log is disabled; our middleware writes one line per request.
|
||||
for name in ("uvicorn", "uvicorn.error"):
|
||||
logging.getLogger(name).handlers[:] = []
|
||||
logging.getLogger(name).propagate = True
|
||||
@@ -0,0 +1,76 @@
|
||||
"""Versioned SQL migrations: `python -m messenger.migrate` (compose service `migrate`).
|
||||
|
||||
* Files `migrations/NNNN_name.sql` are applied in name order, each in its own
|
||||
transaction together with its row in `schema_migrations` (DDL is transactional
|
||||
in PostgreSQL, so a failed migration leaves nothing behind).
|
||||
* Re-running is safe: applied versions are skipped, so data is never duplicated.
|
||||
* The SHA-256 of every applied file is stored; if an applied file was edited later,
|
||||
the run stops instead of silently diverging from the real schema.
|
||||
* `pg_advisory_lock` serialises concurrent runs (two API replicas starting at once).
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import hashlib
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
import psycopg
|
||||
|
||||
from .config import ConfigError, _secret
|
||||
|
||||
MIGRATIONS_DIR = Path(__file__).resolve().parents[2] / "migrations"
|
||||
LOCK_ID = 2026_02 # arbitrary application-wide advisory lock id
|
||||
|
||||
|
||||
def say(text: str) -> None:
|
||||
print(f"migrate: {text}", flush=True)
|
||||
|
||||
|
||||
def connect(url: str, wait_seconds: float = 60) -> psycopg.Connection:
|
||||
deadline = time.monotonic() + wait_seconds
|
||||
while True:
|
||||
try:
|
||||
return psycopg.connect(url, autocommit=True, connect_timeout=3, application_name="messenger-migrate")
|
||||
except psycopg.OperationalError as exc:
|
||||
if time.monotonic() > deadline:
|
||||
raise SystemExit(f"migrate: database is not reachable: {type(exc).__name__}") from None
|
||||
time.sleep(1)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
try:
|
||||
url = _secret("DATABASE_URL", required=True)
|
||||
except ConfigError as exc:
|
||||
print(f"configuration error: {exc}", file=sys.stderr)
|
||||
return 2
|
||||
files = sorted(MIGRATIONS_DIR.glob("*.sql"))
|
||||
with connect(url) as conn:
|
||||
conn.execute("SELECT pg_advisory_lock(%s)", (LOCK_ID,))
|
||||
try:
|
||||
conn.execute(
|
||||
"CREATE TABLE IF NOT EXISTS schema_migrations ("
|
||||
" version text PRIMARY KEY, checksum text NOT NULL,"
|
||||
" applied_at timestamptz NOT NULL DEFAULT now())")
|
||||
applied = dict(conn.execute("SELECT version, checksum FROM schema_migrations").fetchall())
|
||||
for path in files:
|
||||
version, sql = path.stem, path.read_bytes()
|
||||
checksum = hashlib.sha256(sql).hexdigest()
|
||||
if version in applied:
|
||||
if applied[version] != checksum:
|
||||
say(f"{version}: file changed after it was applied; create a new migration instead")
|
||||
return 1
|
||||
say(f"{version}: already applied, skipped")
|
||||
continue
|
||||
with conn.transaction():
|
||||
conn.execute(sql.decode("utf-8"))
|
||||
conn.execute("INSERT INTO schema_migrations (version, checksum) VALUES (%s, %s)", (version, checksum))
|
||||
say(f"{version}: applied")
|
||||
finally:
|
||||
conn.execute("SELECT pg_advisory_unlock(%s)", (LOCK_ID,))
|
||||
say(f"done, {len(files)} migration file(s) checked")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
@@ -0,0 +1,53 @@
|
||||
"""Password hashing with Argon2id (memory-hard password KDF; also mandatory in lab 7).
|
||||
|
||||
* Parameters follow the OWASP minimum for Argon2id: m=19 MiB, t=2, p=1. Unlike bcrypt
|
||||
there is no 72-byte limit, so any 12-128 code point UTF-8 password is hashed whole.
|
||||
* Hashing takes tens of milliseconds of CPU and 19 MiB of RAM. It runs in a worker
|
||||
thread (argon2-cffi releases the GIL), so the event loop keeps serving other
|
||||
requests, and a semaphore caps parallel hashes (memory/CPU DoS protection).
|
||||
* An unknown username is checked against a dummy hash, so "no such user" and
|
||||
"wrong password" take the same time and return the same 401.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
|
||||
from argon2 import PasswordHasher
|
||||
from argon2.exceptions import InvalidHashError, VerificationError, VerifyMismatchError
|
||||
|
||||
_hasher = PasswordHasher(time_cost=2, memory_cost=19 * 1024, parallelism=1, hash_len=32, salt_len=16)
|
||||
_semaphore: asyncio.Semaphore | None = None
|
||||
_dummy_hash: str | None = None
|
||||
|
||||
|
||||
def configure(concurrency: int) -> None:
|
||||
global _semaphore
|
||||
_semaphore = asyncio.Semaphore(concurrency)
|
||||
|
||||
|
||||
async def _run(func, *args):
|
||||
if _semaphore is None:
|
||||
configure(2)
|
||||
async with _semaphore:
|
||||
return await asyncio.to_thread(func, *args)
|
||||
|
||||
|
||||
async def hash_password(password: str) -> str:
|
||||
return await _run(_hasher.hash, password)
|
||||
|
||||
|
||||
def _verify(stored_hash: str, password: str) -> bool:
|
||||
try:
|
||||
return _hasher.verify(stored_hash, password)
|
||||
except (VerifyMismatchError, VerificationError, InvalidHashError):
|
||||
return False
|
||||
|
||||
|
||||
async def verify_password(stored_hash: str | None, password: str) -> bool:
|
||||
global _dummy_hash
|
||||
if stored_hash is None:
|
||||
if _dummy_hash is None:
|
||||
_dummy_hash = await hash_password("dummy-password-for-timing")
|
||||
await _run(_verify, _dummy_hash, password)
|
||||
return False
|
||||
return await _run(_verify, stored_hash, password)
|
||||
@@ -0,0 +1,328 @@
|
||||
"""SQL for users, chats and messages. All statements are parameterised (%s placeholders):
|
||||
values travel separately from the SQL text, so input can never change the query.
|
||||
|
||||
Free text is stored as UTF-8 BYTEA (PostgreSQL TEXT cannot hold U+0000), so it is
|
||||
encoded on the way in and decoded on the way out. See docs/DATA_MODEL.md.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import datetime as dt
|
||||
import uuid
|
||||
from dataclasses import dataclass
|
||||
|
||||
import psycopg
|
||||
from psycopg import errors as pg_errors
|
||||
|
||||
from .cursor import SortKey
|
||||
from .db import Database
|
||||
from .errors import Conflict, InvalidRequest, NotFound, Unavailable
|
||||
from .timefmt import from_us, rfc3339, to_us
|
||||
|
||||
|
||||
def _b(text: str) -> bytes:
|
||||
return text.encode("utf-8")
|
||||
|
||||
|
||||
def _s(raw: bytes | memoryview) -> str:
|
||||
return bytes(raw).decode("utf-8")
|
||||
|
||||
|
||||
@dataclass(slots=True, frozen=True)
|
||||
class User:
|
||||
id: str
|
||||
username: str
|
||||
display_name: str
|
||||
created_at: dt.datetime
|
||||
|
||||
def to_json(self) -> dict:
|
||||
return {"id": self.id, "username": self.username, "display_name": self.display_name,
|
||||
"created_at": rfc3339(self.created_at)}
|
||||
|
||||
|
||||
@dataclass(slots=True, frozen=True)
|
||||
class Chat:
|
||||
id: str
|
||||
title: str
|
||||
member_ids: tuple[str, ...]
|
||||
created_at: dt.datetime
|
||||
|
||||
@property
|
||||
def key(self) -> SortKey:
|
||||
return (to_us(self.created_at), self.id)
|
||||
|
||||
def to_json(self) -> dict:
|
||||
return {"id": self.id, "title": self.title, "member_ids": list(self.member_ids),
|
||||
"created_at": rfc3339(self.created_at)}
|
||||
|
||||
|
||||
@dataclass(slots=True, frozen=True)
|
||||
class Message:
|
||||
id: str
|
||||
chat_id: str
|
||||
sender_id: str
|
||||
text: str
|
||||
created_at: dt.datetime
|
||||
|
||||
@property
|
||||
def key(self) -> SortKey:
|
||||
return (to_us(self.created_at), self.id)
|
||||
|
||||
def to_json(self) -> dict:
|
||||
return {"id": self.id, "chat_id": self.chat_id, "sender_id": self.sender_id, "text": self.text,
|
||||
"created_at": rfc3339(self.created_at)}
|
||||
|
||||
|
||||
def _user(row) -> User:
|
||||
return User(str(row[0]), row[1], _s(row[2]), row[3])
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ users
|
||||
async def create_user(db: Database, username: str, display_name: str, password_hash: str) -> User:
|
||||
async def work(conn: psycopg.AsyncConnection) -> User:
|
||||
cur = await conn.execute(
|
||||
"INSERT INTO users (id, username, display_name, password_hash) VALUES (%s, %s, %s, %s) "
|
||||
"RETURNING id, username, display_name, created_at",
|
||||
(uuid.uuid4(), username, _b(display_name), password_hash))
|
||||
return _user(await cur.fetchone())
|
||||
try:
|
||||
return await db.run(work)
|
||||
except pg_errors.UniqueViolation:
|
||||
# Two concurrent registrations: the unique index lets exactly one INSERT commit.
|
||||
raise Conflict("Имя пользователя уже занято") from None
|
||||
|
||||
|
||||
async def get_user(db: Database, user_id: str) -> User | None:
|
||||
async def work(conn: psycopg.AsyncConnection) -> User | None:
|
||||
cur = await conn.execute(
|
||||
"SELECT id, username, display_name, created_at FROM users WHERE id = %s", (user_id,))
|
||||
row = await cur.fetchone()
|
||||
return _user(row) if row else None
|
||||
return await db.run(work)
|
||||
|
||||
|
||||
async def get_credentials(db: Database, username: str) -> tuple[User, str] | None:
|
||||
async def work(conn: psycopg.AsyncConnection) -> tuple[User, str] | None:
|
||||
cur = await conn.execute(
|
||||
"SELECT id, username, display_name, created_at, password_hash FROM users WHERE username = %s",
|
||||
(username,))
|
||||
row = await cur.fetchone()
|
||||
return (_user(row), row[4]) if row else None
|
||||
return await db.run(work)
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ chats
|
||||
async def create_chat(db: Database, creator_id: str, title: str, member_ids: list[str]) -> Chat:
|
||||
"""Atomic: the chat row and all memberships commit together or not at all.
|
||||
|
||||
The chat is inserted FIRST; an unknown member then fails the membership INSERT
|
||||
with a foreign-key violation and the whole transaction (chat row included) is
|
||||
rolled back. student-tests check that nothing is left behind.
|
||||
"""
|
||||
if creator_id not in member_ids:
|
||||
raise InvalidRequest("member_ids должен включать текущего пользователя")
|
||||
chat_id = uuid.uuid4()
|
||||
|
||||
async def work(conn: psycopg.AsyncConnection) -> dt.datetime:
|
||||
cur = await conn.execute(
|
||||
"INSERT INTO chats (id, title, created_by) VALUES (%s, %s, %s) RETURNING created_at",
|
||||
(chat_id, _b(title), creator_id))
|
||||
(created_at,) = await cur.fetchone()
|
||||
await conn.execute(
|
||||
"INSERT INTO chat_members (chat_id, user_id, position, chat_created_at) "
|
||||
"SELECT %s, m.user_id, m.ord - 1, %s "
|
||||
"FROM unnest(%s::uuid[]) WITH ORDINALITY AS m(user_id, ord)",
|
||||
(chat_id, created_at, member_ids))
|
||||
return created_at
|
||||
try:
|
||||
created_at = await db.run(work)
|
||||
except pg_errors.ForeignKeyViolation:
|
||||
raise InvalidRequest("member_ids содержит несуществующих пользователей") from None
|
||||
return Chat(str(chat_id), title, tuple(member_ids), created_at)
|
||||
|
||||
|
||||
async def _load_chats(conn: psycopg.AsyncConnection, rows) -> list[Chat]:
|
||||
if not rows:
|
||||
return []
|
||||
ids = [r[0] for r in rows]
|
||||
cur = await conn.execute(
|
||||
"SELECT chat_id, user_id FROM chat_members WHERE chat_id = ANY(%s) ORDER BY chat_id, position", (ids,))
|
||||
members: dict[uuid.UUID, list[str]] = {i: [] for i in ids}
|
||||
for chat_id, user_id in await cur.fetchall():
|
||||
members[chat_id].append(str(user_id))
|
||||
return [Chat(str(r[0]), _s(r[1]), tuple(members[r[0]]), r[2]) for r in rows]
|
||||
|
||||
|
||||
async def get_chat(db: Database, user_id: str, chat_id: str) -> Chat:
|
||||
"""The chat if the user is a member; a missing chat and a foreign chat are both 404."""
|
||||
async def work(conn: psycopg.AsyncConnection) -> Chat:
|
||||
cur = await conn.execute(
|
||||
"SELECT c.id, c.title, c.created_at FROM chats c "
|
||||
"JOIN chat_members m ON m.chat_id = c.id AND m.user_id = %s WHERE c.id = %s",
|
||||
(user_id, chat_id))
|
||||
row = await cur.fetchone()
|
||||
if row is None:
|
||||
raise NotFound("Чат не найден")
|
||||
return (await _load_chats(conn, [row]))[0]
|
||||
return await db.run(work)
|
||||
|
||||
|
||||
async def _check_member(conn: psycopg.AsyncConnection, user_id: str, chat_id: str) -> None:
|
||||
cur = await conn.execute(
|
||||
"SELECT 1 FROM chat_members WHERE chat_id = %s AND user_id = %s", (chat_id, user_id))
|
||||
if await cur.fetchone() is None:
|
||||
raise NotFound("Чат не найден")
|
||||
|
||||
|
||||
async def ensure_member(db: Database, user_id: str, chat_id: str) -> None:
|
||||
async def work(conn: psycopg.AsyncConnection) -> None:
|
||||
await _check_member(conn, user_id, chat_id)
|
||||
await db.run(work)
|
||||
|
||||
|
||||
async def list_chats(db: Database, user_id: str, after: SortKey | None, limit: int) -> tuple[list[Chat], bool]:
|
||||
async def work(conn: psycopg.AsyncConnection) -> tuple[list[Chat], bool]:
|
||||
if after is None:
|
||||
cur = await conn.execute(
|
||||
"SELECT c.id, c.title, c.created_at FROM chat_members m JOIN chats c ON c.id = m.chat_id "
|
||||
"WHERE m.user_id = %s ORDER BY m.chat_created_at, m.chat_id LIMIT %s",
|
||||
(user_id, limit + 1))
|
||||
else:
|
||||
cur = await conn.execute(
|
||||
"SELECT c.id, c.title, c.created_at FROM chat_members m JOIN chats c ON c.id = m.chat_id "
|
||||
"WHERE m.user_id = %s AND (m.chat_created_at, m.chat_id) > (%s, %s) "
|
||||
"ORDER BY m.chat_created_at, m.chat_id LIMIT %s",
|
||||
(user_id, from_us(after[0]), after[1], limit + 1))
|
||||
rows = await cur.fetchall()
|
||||
return await _load_chats(conn, rows[:limit]), len(rows) > limit
|
||||
return await db.run(work)
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ messages
|
||||
@dataclass(slots=True, frozen=True)
|
||||
class PostResult:
|
||||
message: Message
|
||||
replayed: bool # True: the key was already used with this body, nothing new was written
|
||||
|
||||
|
||||
async def _insert_message(conn: psycopg.AsyncConnection, message_id: uuid.UUID, user_id: str, chat_id: str,
|
||||
text: str) -> dt.datetime:
|
||||
"""Insert with a per-chat strictly increasing timestamp.
|
||||
|
||||
`UPDATE chats ... RETURNING` locks the chat row until COMMIT, so writers of one
|
||||
chat are serialised: a message with a smaller created_at can never commit after
|
||||
a message with a bigger one, and a reader paging by cursor cannot skip it.
|
||||
"""
|
||||
cur = await conn.execute(
|
||||
"UPDATE chats SET last_message_at = greatest(clock_timestamp(), "
|
||||
"last_message_at + interval '1 microsecond') WHERE id = %s RETURNING last_message_at",
|
||||
(chat_id,))
|
||||
(created_at,) = await cur.fetchone()
|
||||
await conn.execute(
|
||||
"INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES (%s, %s, %s, %s, %s)",
|
||||
(message_id, chat_id, user_id, _b(text), created_at))
|
||||
return created_at
|
||||
|
||||
|
||||
# Claim the key. A row of a concurrent transaction that is not committed yet makes this
|
||||
# INSERT wait on the unique index until that transaction ends: after COMMIT it is a conflict
|
||||
# (the key is taken), after ROLLBACK the INSERT goes through. A key older than the TTL is
|
||||
# re-used as if it were new (ON CONFLICT ... DO UPDATE ... WHERE expired).
|
||||
CLAIM_KEY = (
|
||||
"INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) "
|
||||
"VALUES (%s, %s, %s, %s, %s) "
|
||||
"ON CONFLICT (user_id, chat_id, key) DO UPDATE SET "
|
||||
"request_hash = EXCLUDED.request_hash, message_id = EXCLUDED.message_id, created_at = clock_timestamp() "
|
||||
"WHERE idempotency_keys.created_at < clock_timestamp() - make_interval(secs => %s) "
|
||||
"RETURNING 1")
|
||||
STORED_RESULT = (
|
||||
"SELECT k.request_hash, m.id, m.text, m.created_at FROM idempotency_keys k "
|
||||
"JOIN messages m ON m.id = k.message_id "
|
||||
"WHERE k.user_id = %s AND k.chat_id = %s AND k.key = %s")
|
||||
|
||||
|
||||
async def post_message(db: Database, user_id: str, chat_id: str, key: str, request_hash: bytes, text: str,
|
||||
ttl_seconds: int) -> PostResult:
|
||||
"""Idempotent send: the key, the message and the stored result commit in ONE transaction.
|
||||
|
||||
* first request -> key row + message + chat clock, all or nothing (no "pending" key is
|
||||
ever left behind: a crash or 5xx rolls everything back and the client may retry);
|
||||
* same key, same body (any replica, also concurrently) -> the stored Message, 201, replayed;
|
||||
* same key, other body -> 409; the membership check comes first, so outsiders always get 404.
|
||||
"""
|
||||
message_id = uuid.uuid4()
|
||||
|
||||
async def work(conn: psycopg.AsyncConnection) -> PostResult:
|
||||
await _check_member(conn, user_id, chat_id)
|
||||
cur = await conn.execute(CLAIM_KEY, (user_id, chat_id, key, request_hash, message_id, ttl_seconds))
|
||||
if await cur.fetchone() is not None:
|
||||
created_at = await _insert_message(conn, message_id, user_id, chat_id, text)
|
||||
return PostResult(Message(str(message_id), chat_id, user_id, text, created_at), False)
|
||||
cur = await conn.execute(STORED_RESULT, (user_id, chat_id, key))
|
||||
row = await cur.fetchone()
|
||||
if row is None:
|
||||
# The key row vanished between the two statements (expired-key cleanup): retryable.
|
||||
raise Unavailable("Повторите запрос с тем же Idempotency-Key")
|
||||
if bytes(row[0]) != request_hash:
|
||||
raise Conflict("Idempotency-Key уже использован с другим сообщением")
|
||||
return PostResult(Message(str(row[1]), chat_id, user_id, _s(row[2]), row[3]), True)
|
||||
return await db.run(work)
|
||||
|
||||
|
||||
POST_MESSAGE_CALL = "SELECT o_status, o_id, o_text, o_created_at FROM post_message(%s, %s, %s, %s, %s, %s, %s)"
|
||||
|
||||
|
||||
async def post_message_single_call(db: Database, user_id: str, chat_id: str, key: str, request_hash: bytes,
|
||||
text: str, ttl_seconds: int) -> PostResult:
|
||||
"""Same contract as post_message, ONE round trip: SQL function post_message (migration 0003)
|
||||
in autocommit mode, i.e. the call is the whole transaction. Track 1 improvement."""
|
||||
message_id = uuid.uuid4()
|
||||
|
||||
async def work(conn: psycopg.AsyncConnection) -> tuple:
|
||||
cur = await conn.execute(POST_MESSAGE_CALL,
|
||||
(user_id, chat_id, key, request_hash, message_id, _b(text), ttl_seconds))
|
||||
return await cur.fetchone()
|
||||
status, stored_id, stored_text, created_at = await db.run(work, transaction=False)
|
||||
if status == "not_found":
|
||||
raise NotFound("Чат не найден")
|
||||
if status == "conflict":
|
||||
raise Conflict("Idempotency-Key уже использован с другим сообщением")
|
||||
if status == "retry":
|
||||
raise Unavailable("Повторите запрос с тем же Idempotency-Key")
|
||||
return PostResult(Message(str(stored_id), chat_id, user_id, _s(stored_text), created_at), status == "replayed")
|
||||
|
||||
|
||||
async def purge_idempotency_keys(db: Database, ttl_seconds: int, batch: int = 5000) -> int:
|
||||
"""Delete expired keys in bounded batches; one replica at a time (advisory xact lock)."""
|
||||
async def work(conn: psycopg.AsyncConnection) -> int:
|
||||
cur = await conn.execute("SELECT pg_try_advisory_xact_lock(%s)", (2026_03,))
|
||||
if not (await cur.fetchone())[0]:
|
||||
return 0
|
||||
cur = await conn.execute(
|
||||
"DELETE FROM idempotency_keys WHERE ctid = ANY(ARRAY("
|
||||
"SELECT ctid FROM idempotency_keys "
|
||||
"WHERE created_at < clock_timestamp() - make_interval(secs => %s) LIMIT %s))",
|
||||
(ttl_seconds, batch))
|
||||
return cur.rowcount
|
||||
return await db.run(work)
|
||||
|
||||
|
||||
# Two statements instead of one with "(%s IS NULL OR ...)": with a generic plan the OR
|
||||
# would push the cursor into a Filter and late pages would scan the whole prefix.
|
||||
HISTORY_FIRST = (
|
||||
"SELECT id, sender_id, text, created_at FROM messages "
|
||||
"WHERE chat_id = %s ORDER BY created_at, id LIMIT %s")
|
||||
HISTORY_AFTER = (
|
||||
"SELECT id, sender_id, text, created_at FROM messages "
|
||||
"WHERE chat_id = %s AND (created_at, id) > (%s, %s) ORDER BY created_at, id LIMIT %s")
|
||||
|
||||
|
||||
async def list_messages(db: Database, chat_id: str, after: SortKey | None, limit: int) -> tuple[list[Message], bool]:
|
||||
async def work(conn: psycopg.AsyncConnection) -> list:
|
||||
if after is None:
|
||||
cur = await conn.execute(HISTORY_FIRST, (chat_id, limit + 1))
|
||||
else:
|
||||
cur = await conn.execute(HISTORY_AFTER, (chat_id, from_us(after[0]), after[1], limit + 1))
|
||||
return await cur.fetchall()
|
||||
rows = await db.run(work)
|
||||
items = [Message(str(r[0]), chat_id, str(r[1]), _s(r[2]), r[3]) for r in rows[:limit]]
|
||||
return items, len(rows) > limit
|
||||
@@ -0,0 +1,160 @@
|
||||
"""Server-side sessions behind an opaque Bearer token.
|
||||
|
||||
* Token: `secrets.token_urlsafe(32)` = 256 random bits. Only SHA-256(token) is
|
||||
stored, so a leaked database/AOF does not contain usable tokens. A fast hash is
|
||||
enough for a 256-bit random value; passwords need a slow KDF (passwords.py).
|
||||
* Expiry is absolute: `expires_at = login time + SESSION_TTL_SECONDS`. Every lookup
|
||||
compares it with the current time, and a restart never extends it.
|
||||
|
||||
Two interchangeable stores (SESSION_STORE):
|
||||
|
||||
* `valkey` (compose default): the ONLY copy of a session is a Valkey key with
|
||||
`PXAT expires_at`, persisted by AOF with `appendfsync always`. There is no second
|
||||
copy (no cache over PostgreSQL), so a logout cannot be "forgotten" and a revoked
|
||||
token cannot come back after Valkey fails and recovers. If Valkey is unreachable
|
||||
a session cannot be checked: protected requests and logout get 503, never access.
|
||||
* `postgres`: table `sessions` (migrations/0001_init.sql).
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import datetime as dt
|
||||
import hashlib
|
||||
import json
|
||||
import logging
|
||||
import secrets
|
||||
from typing import Protocol
|
||||
|
||||
from redis import exceptions as redis_errors
|
||||
from redis.asyncio import Redis
|
||||
from redis.backoff import NoBackoff
|
||||
from redis.retry import Retry
|
||||
|
||||
from .config import Settings
|
||||
from .db import Database
|
||||
from .errors import Unavailable
|
||||
from .timefmt import UTC, from_us, to_us
|
||||
|
||||
log = logging.getLogger("messenger.sessions")
|
||||
|
||||
|
||||
def new_token() -> str:
|
||||
return secrets.token_urlsafe(32)
|
||||
|
||||
|
||||
def token_hash(token: str) -> bytes:
|
||||
return hashlib.sha256(token.encode("ascii")).digest()
|
||||
|
||||
|
||||
def now() -> dt.datetime:
|
||||
return dt.datetime.now(UTC)
|
||||
|
||||
|
||||
class SessionStore(Protocol):
|
||||
name: str
|
||||
|
||||
async def create(self, user_id: str, ttl_seconds: int) -> tuple[str, dt.datetime]: ...
|
||||
async def user_id_for(self, token: str) -> str | None: ...
|
||||
async def revoke(self, token: str) -> bool: ...
|
||||
async def ping(self) -> bool: ...
|
||||
async def close(self) -> None: ...
|
||||
|
||||
|
||||
class PostgresSessionStore:
|
||||
name = "postgres"
|
||||
|
||||
def __init__(self, db: Database):
|
||||
self.db = db
|
||||
|
||||
async def create(self, user_id: str, ttl_seconds: int) -> tuple[str, dt.datetime]:
|
||||
token = new_token()
|
||||
expires_at = now() + dt.timedelta(seconds=ttl_seconds)
|
||||
|
||||
async def work(conn) -> None:
|
||||
await conn.execute(
|
||||
"INSERT INTO sessions (token_hash, user_id, expires_at) VALUES (%s, %s, %s)",
|
||||
(token_hash(token), user_id, expires_at))
|
||||
# Housekeeping: drop this user's expired sessions (bounded by the user's own rows).
|
||||
await conn.execute("DELETE FROM sessions WHERE user_id = %s AND expires_at <= %s", (user_id, now()))
|
||||
await self.db.run(work)
|
||||
return token, expires_at
|
||||
|
||||
async def user_id_for(self, token: str) -> str | None:
|
||||
async def work(conn):
|
||||
cur = await conn.execute(
|
||||
"SELECT user_id FROM sessions WHERE token_hash = %s AND expires_at > %s", (token_hash(token), now()))
|
||||
return await cur.fetchone()
|
||||
row = await self.db.run(work)
|
||||
return str(row[0]) if row else None
|
||||
|
||||
async def revoke(self, token: str) -> bool:
|
||||
async def work(conn) -> bool:
|
||||
cur = await conn.execute(
|
||||
"DELETE FROM sessions WHERE token_hash = %s AND expires_at > %s", (token_hash(token), now()))
|
||||
return cur.rowcount == 1
|
||||
return await self.db.run(work)
|
||||
|
||||
async def ping(self) -> bool:
|
||||
return True # covered by the database readiness check
|
||||
|
||||
async def close(self) -> None:
|
||||
return None
|
||||
|
||||
|
||||
class ValkeySessionStore:
|
||||
name = "valkey"
|
||||
PREFIX = "session:"
|
||||
|
||||
def __init__(self, settings: Settings):
|
||||
self.timeout = settings.cache_timeout_seconds
|
||||
self.client = Redis.from_url(
|
||||
settings.cache_url,
|
||||
socket_timeout=self.timeout,
|
||||
socket_connect_timeout=self.timeout,
|
||||
retry=Retry(NoBackoff(), 0), # no hidden retries: a failure is reported at once
|
||||
decode_responses=False,
|
||||
)
|
||||
|
||||
def _key(self, token: str) -> str:
|
||||
return self.PREFIX + token_hash(token).hex()
|
||||
|
||||
async def _call(self, coro):
|
||||
try:
|
||||
async with asyncio.timeout(self.timeout * 2):
|
||||
return await coro
|
||||
except (redis_errors.ConnectionError, redis_errors.TimeoutError, TimeoutError, OSError) as exc:
|
||||
log.warning("session_store_unavailable", extra={"fields": {"error": type(exc).__name__}})
|
||||
raise Unavailable("Хранилище сессий временно недоступно") from None
|
||||
|
||||
async def create(self, user_id: str, ttl_seconds: int) -> tuple[str, dt.datetime]:
|
||||
token = new_token()
|
||||
expires_at = now() + dt.timedelta(seconds=ttl_seconds)
|
||||
value = json.dumps({"user_id": user_id, "expires_at_us": to_us(expires_at)}).encode()
|
||||
# PXAT = absolute expiry time: survives restarts (AOF) without being extended.
|
||||
await self._call(self.client.set(self._key(token), value, pxat=to_us(expires_at) // 1000))
|
||||
return token, expires_at
|
||||
|
||||
async def user_id_for(self, token: str) -> str | None:
|
||||
raw = await self._call(self.client.get(self._key(token)))
|
||||
if raw is None:
|
||||
return None
|
||||
data = json.loads(raw)
|
||||
if from_us(data["expires_at_us"]) <= now():
|
||||
return None # do not rely only on Valkey's lazy expiry
|
||||
return data["user_id"]
|
||||
|
||||
async def revoke(self, token: str) -> bool:
|
||||
return await self._call(self.client.delete(self._key(token))) == 1
|
||||
|
||||
async def ping(self) -> bool:
|
||||
try:
|
||||
return bool(await self._call(self.client.ping()))
|
||||
except Unavailable:
|
||||
return False
|
||||
|
||||
async def close(self) -> None:
|
||||
await self.client.aclose()
|
||||
|
||||
|
||||
def build_store(settings: Settings, db: Database) -> SessionStore:
|
||||
return ValkeySessionStore(settings) if settings.session_store == "valkey" else PostgresSessionStore(db)
|
||||
@@ -0,0 +1,22 @@
|
||||
"""Conversions between PostgreSQL timestamptz values, cursor integers and RFC 3339 strings."""
|
||||
from __future__ import annotations
|
||||
|
||||
import datetime as dt
|
||||
|
||||
UTC = dt.timezone.utc
|
||||
_EPOCH = dt.datetime(1970, 1, 1, tzinfo=UTC)
|
||||
|
||||
|
||||
def to_us(value: dt.datetime) -> int:
|
||||
"""Microseconds since the Unix epoch (exact: timestamptz has microsecond precision)."""
|
||||
delta = value - _EPOCH
|
||||
return (delta.days * 86_400 + delta.seconds) * 1_000_000 + delta.microseconds
|
||||
|
||||
|
||||
def from_us(us: int) -> dt.datetime:
|
||||
return _EPOCH + dt.timedelta(microseconds=us)
|
||||
|
||||
|
||||
def rfc3339(value: dt.datetime) -> str:
|
||||
"""RFC 3339 UTC with fixed microsecond precision, e.g. 2026-10-12T12:00:00.000001Z."""
|
||||
return value.astimezone(UTC).strftime("%Y-%m-%dT%H:%M:%S.%fZ")
|
||||
@@ -0,0 +1,147 @@
|
||||
"""Strict request validation according to contracts/openapi.json + contracts/README.md.
|
||||
|
||||
Written by hand instead of a schema library so that every rule is visible:
|
||||
extra fields, wrong JSON types, whitespace-only strings and lengths counted in
|
||||
Unicode code points all produce 400 invalid_request.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from typing import Any
|
||||
|
||||
from .errors import InvalidRequest
|
||||
|
||||
USERNAME_RE = re.compile(r"[a-z0-9_]{3,32}")
|
||||
# Canonical textual UUID (8-4-4-4-12 hex). Case-insensitive, normalised to lower case.
|
||||
UUID_RE = re.compile(r"[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}")
|
||||
LIMIT_RE = re.compile(r"[0-9]{1,3}")
|
||||
# Idempotency-Key (contract, lab 3): ASCII [A-Za-z0-9._:-]{1,128}.
|
||||
IDEMPOTENCY_KEY_RE = re.compile(r"[A-Za-z0-9._:-]{1,128}")
|
||||
SURROGATES_RE = re.compile("[\ud800-\udfff]")
|
||||
|
||||
DEFAULT_LIMIT = 50
|
||||
MAX_LIMIT = 100
|
||||
|
||||
|
||||
def _reject_constant(name: str) -> Any:
|
||||
# Python's json accepts NaN/Infinity, which are not valid JSON.
|
||||
raise ValueError(f"invalid JSON constant {name}")
|
||||
|
||||
|
||||
def _unique_keys(pairs: list[tuple[str, Any]]) -> dict[str, Any]:
|
||||
# {"text":"a","text":"b"} is ambiguous (which one is meant?) -> 400, not "last one wins".
|
||||
result = dict(pairs)
|
||||
if len(result) != len(pairs):
|
||||
raise ValueError("duplicate key in JSON object")
|
||||
return result
|
||||
|
||||
|
||||
def parse_json_object(body: bytes) -> dict[str, Any]:
|
||||
"""Decode a UTF-8 JSON object or raise InvalidRequest (malformed JSON -> 400)."""
|
||||
try:
|
||||
text = body.decode("utf-8")
|
||||
value = json.loads(text, parse_constant=_reject_constant, object_pairs_hook=_unique_keys)
|
||||
except (UnicodeDecodeError, ValueError):
|
||||
raise InvalidRequest("Тело запроса должно быть корректным JSON в UTF-8") from None
|
||||
if not isinstance(value, dict):
|
||||
raise InvalidRequest("Тело запроса должно быть JSON-объектом")
|
||||
return value
|
||||
|
||||
|
||||
def check_fields(data: dict[str, Any], required: set[str], optional: frozenset[str] = frozenset()) -> None:
|
||||
extra = set(data) - required - optional
|
||||
if extra:
|
||||
raise InvalidRequest("Неизвестные поля: " + ", ".join(sorted(extra)))
|
||||
missing = required - set(data)
|
||||
if missing:
|
||||
raise InvalidRequest("Отсутствуют обязательные поля: " + ", ".join(sorted(missing)))
|
||||
|
||||
|
||||
def text_field(data: dict[str, Any], name: str, max_len: int, *, min_len: int = 1) -> str:
|
||||
value = data[name]
|
||||
if not isinstance(value, str):
|
||||
raise InvalidRequest(f"Поле {name} должно быть строкой")
|
||||
if SURROGATES_RE.search(value):
|
||||
# "\ud800" escapes decode to lone surrogates that cannot be encoded as UTF-8.
|
||||
raise InvalidRequest(f"Поле {name} содержит недопустимые символы")
|
||||
# len() of a Python str is the number of Unicode code points, as the contract requires.
|
||||
if not min_len <= len(value) <= max_len:
|
||||
raise InvalidRequest(f"Длина поля {name} должна быть от {min_len} до {max_len} символов")
|
||||
if value.strip() == "":
|
||||
raise InvalidRequest(f"Поле {name} не может состоять только из пробелов")
|
||||
return value
|
||||
|
||||
|
||||
def username_field(data: dict[str, Any]) -> str:
|
||||
value = data["username"]
|
||||
if not isinstance(value, str) or USERNAME_RE.fullmatch(value) is None:
|
||||
raise InvalidRequest("username должен соответствовать [a-z0-9_]{3,32}")
|
||||
return value
|
||||
|
||||
|
||||
def password_field(data: dict[str, Any]) -> str:
|
||||
"""12-128 Unicode code points, not whitespace-only; any other character (':' too) is allowed."""
|
||||
value = data["password"]
|
||||
if not isinstance(value, str):
|
||||
raise InvalidRequest("Поле password должно быть строкой")
|
||||
if SURROGATES_RE.search(value):
|
||||
raise InvalidRequest("Поле password содержит недопустимые символы")
|
||||
if not 12 <= len(value) <= 128:
|
||||
raise InvalidRequest("Длина пароля должна быть от 12 до 128 символов")
|
||||
if value.strip() == "":
|
||||
raise InvalidRequest("Пароль не может состоять только из пробелов")
|
||||
return value
|
||||
|
||||
|
||||
def parse_uuid(value: Any, what: str = "UUID") -> str:
|
||||
if not isinstance(value, str) or UUID_RE.fullmatch(value) is None:
|
||||
raise InvalidRequest(f"Неверный {what}")
|
||||
return value.lower()
|
||||
|
||||
|
||||
def member_ids_field(data: dict[str, Any]) -> list[str]:
|
||||
value = data["member_ids"]
|
||||
if not isinstance(value, list):
|
||||
raise InvalidRequest("member_ids должен быть массивом")
|
||||
if not 1 <= len(value) <= 100:
|
||||
raise InvalidRequest("member_ids должен содержать от 1 до 100 элементов")
|
||||
ids = [parse_uuid(item, "UUID участника") for item in value]
|
||||
if len(set(ids)) != len(ids):
|
||||
raise InvalidRequest("member_ids содержит дубликаты")
|
||||
return ids
|
||||
|
||||
|
||||
def attachment_ids_field(data: dict[str, Any]) -> list[str]:
|
||||
"""Optional `attachment_ids` (lab 4+). Absent and [] are the same message (contracts/README.md);
|
||||
lab 3 has no attachments yet, so a non-empty list can only reference unknown objects."""
|
||||
value = data.get("attachment_ids", [])
|
||||
if not isinstance(value, list):
|
||||
raise InvalidRequest("attachment_ids должен быть массивом")
|
||||
if len(value) > 10:
|
||||
raise InvalidRequest("attachment_ids содержит больше 10 элементов")
|
||||
ids = [parse_uuid(item, "UUID вложения") for item in value]
|
||||
if len(set(ids)) != len(ids):
|
||||
raise InvalidRequest("attachment_ids содержит дубликаты")
|
||||
return ids
|
||||
|
||||
|
||||
def idempotency_key(raw: str | None) -> str:
|
||||
if raw is None or IDEMPOTENCY_KEY_RE.fullmatch(raw) is None:
|
||||
raise InvalidRequest("Заголовок Idempotency-Key обязателен: [A-Za-z0-9._:-]{1,128}")
|
||||
return raw
|
||||
|
||||
|
||||
def single_query_param(query: list[tuple[str, str]], name: str) -> str | None:
|
||||
values = [v for k, v in query if k == name]
|
||||
if len(values) > 1:
|
||||
raise InvalidRequest(f"Параметр {name} указан несколько раз")
|
||||
return values[0] if values else None
|
||||
|
||||
|
||||
def parse_limit(raw: str | None) -> int:
|
||||
if raw is None:
|
||||
return DEFAULT_LIMIT
|
||||
if LIMIT_RE.fullmatch(raw) is None or not 1 <= int(raw) <= MAX_LIMIT:
|
||||
raise InvalidRequest(f"limit должен быть целым числом от 1 до {MAX_LIMIT}")
|
||||
return int(raw)
|
||||
@@ -0,0 +1,291 @@
|
||||
"""Shared helpers for the student tests.
|
||||
|
||||
* `unit/` — pure functions (validation, cursor, auth headers, Argon2id); no services.
|
||||
* `test_*` in this folder — marker `stack`: black-box HTTP against the running compose
|
||||
stand THROUGH THE LOAD BALANCER (BASE_URL, default http://127.0.0.1:8080) plus
|
||||
`docker compose` for database checks (psql inside the postgres container), the HAProxy
|
||||
runtime state (scripts/lbctl.py) and failure injection. Every failure test restores the
|
||||
stopped/paused service or replica in `finally` and waits until HAProxy sees all replicas UP.
|
||||
|
||||
python -m pytest student-tests # everything (stand must be up)
|
||||
python -m pytest student-tests -m "not stack" # unit only
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import contextlib
|
||||
import json
|
||||
import os
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
import uuid
|
||||
from pathlib import Path
|
||||
|
||||
import httpx2 as httpx
|
||||
import pytest
|
||||
from jsonschema import Draft202012Validator
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
sys.path.insert(0, str(ROOT / "src"))
|
||||
sys.path.insert(0, str(ROOT / "scripts"))
|
||||
import lbctl # noqa: E402
|
||||
|
||||
SPEC = json.loads((ROOT / "contracts" / "openapi.json").read_text(encoding="utf-8"))
|
||||
BASE_URL = os.environ.get("BASE_URL", "http://127.0.0.1:8080").rstrip("/")
|
||||
PASSWORD = "Test-password-Ω-12345" # synthetic; 12+ code points, non-ASCII on purpose
|
||||
REPLICAS = int(os.environ.get("API_REPLICAS", "2")) # replicas the stand normally runs
|
||||
|
||||
|
||||
def pytest_configure(config):
|
||||
config.addinivalue_line("markers", "stack: needs the running compose stand (docker compose up -d --wait)")
|
||||
|
||||
|
||||
def schema_errors(value, name: str) -> list[str]:
|
||||
"""Validate against components/schemas/<name> of the course contract (JSON Schema 2020-12)."""
|
||||
schema = {"$ref": f"#/components/schemas/{name}", "components": SPEC["components"]}
|
||||
validator = Draft202012Validator(schema, format_checker=Draft202012Validator.FORMAT_CHECKER)
|
||||
return [e.message for e in validator.iter_errors(value)]
|
||||
|
||||
|
||||
def basic(username: str, password: str) -> str:
|
||||
return "Basic " + base64.b64encode(f"{username}:{password}".encode()).decode()
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ docker / psql
|
||||
def compose(*args: str, check: bool = True, timeout: float = 120) -> subprocess.CompletedProcess:
|
||||
return subprocess.run(["docker", "compose", *args], cwd=ROOT, capture_output=True, text=True,
|
||||
encoding="utf-8", check=check, timeout=timeout)
|
||||
|
||||
|
||||
def psql(sql: str) -> str:
|
||||
"""Run SQL inside the postgres container (no DB port is published) and return unaligned output."""
|
||||
result = compose("exec", "-T", "postgres", "psql", "-U", "messenger", "-d", "messenger",
|
||||
"-At", "-v", "ON_ERROR_STOP=1", "-c", sql)
|
||||
return result.stdout.strip()
|
||||
|
||||
|
||||
def wait_ready(timeout: float = 90) -> None:
|
||||
deadline = time.monotonic() + timeout
|
||||
while time.monotonic() < deadline:
|
||||
try:
|
||||
if httpx.get(BASE_URL + "/health/ready", timeout=2).status_code == 200:
|
||||
return
|
||||
except httpx.HTTPError:
|
||||
pass
|
||||
time.sleep(0.3)
|
||||
raise AssertionError("stand did not become ready")
|
||||
|
||||
|
||||
def wait_healthy(*services: str, timeout: float = 120) -> None:
|
||||
"""Poll Docker health until EVERY container of every service is healthy again. `compose up
|
||||
--wait` is not used: it gives up at once on a container marked unhealthy while paused."""
|
||||
deadline = time.monotonic() + timeout
|
||||
pending = set(services)
|
||||
while pending and time.monotonic() < deadline:
|
||||
for service in list(pending):
|
||||
ids = compose("ps", "-q", service, check=False).stdout.split()
|
||||
if ids:
|
||||
states = subprocess.run(["docker", "inspect", "-f", "{{.State.Health.Status}}", *ids],
|
||||
capture_output=True, text=True).stdout.split()
|
||||
if states and all(state == "healthy" for state in states):
|
||||
pending.discard(service)
|
||||
time.sleep(0.5)
|
||||
assert not pending, f"not healthy: {pending}"
|
||||
|
||||
|
||||
def wait_stand(replicas: int = REPLICAS) -> None:
|
||||
"""Readiness through the LB AND every replica back in HAProxy rotation."""
|
||||
wait_ready()
|
||||
lbctl.wait_up(replicas, timeout=90)
|
||||
|
||||
|
||||
def restart_services(*services: str) -> None:
|
||||
compose("restart", *services)
|
||||
wait_healthy(*services)
|
||||
wait_stand()
|
||||
|
||||
|
||||
@contextlib.contextmanager
|
||||
def service_down(service: str, how: str = "stop"):
|
||||
"""Stop or pause a compose service; ALWAYS bring it back and wait for readiness."""
|
||||
assert how in {"stop", "pause"}
|
||||
compose(how, service)
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
compose("unpause" if how == "pause" else "start", service)
|
||||
wait_healthy(service)
|
||||
wait_stand()
|
||||
|
||||
|
||||
@contextlib.contextmanager
|
||||
def api_with_env(**overrides: str):
|
||||
"""Recreate the api container with environment overrides (e.g. SESSION_TTL_SECONDS=2),
|
||||
then ALWAYS recreate it with the normal configuration."""
|
||||
env = {**os.environ, **overrides}
|
||||
subprocess.run(["docker", "compose", "up", "-d", "--wait", "api"], cwd=ROOT, env=env,
|
||||
capture_output=True, text=True, check=True, timeout=180)
|
||||
wait_stand()
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
clean = {k: v for k, v in os.environ.items() if k not in overrides}
|
||||
subprocess.run(["docker", "compose", "up", "-d", "--wait", "api"], cwd=ROOT, env=clean,
|
||||
capture_output=True, text=True, check=True, timeout=180)
|
||||
wait_stand()
|
||||
|
||||
|
||||
def running_replicas() -> list[dict]:
|
||||
return [r for r in lbctl.replicas() if r["state"] == "running"]
|
||||
|
||||
|
||||
@contextlib.contextmanager
|
||||
def replica_stopped(replica: dict, mode: str = "stop"):
|
||||
"""Stop ONE api container (stop = SIGTERM + drain, kill = SIGKILL); ALWAYS start it again
|
||||
and wait until HAProxy has every replica UP."""
|
||||
assert mode in {"stop", "kill"}
|
||||
subprocess.run(["docker", "stop", "--time", "15" if mode == "stop" else "0", replica["id"]],
|
||||
capture_output=True, text=True, check=True, timeout=60)
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
subprocess.run(["docker", "start", replica["id"]], capture_output=True, text=True, check=True, timeout=60)
|
||||
wait_healthy("api")
|
||||
wait_stand()
|
||||
|
||||
|
||||
def api_env(name: str) -> str:
|
||||
"""Current value of an environment variable inside the running api container."""
|
||||
return compose("exec", "-T", "api", "printenv", name).stdout.strip()
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ HTTP
|
||||
class Api:
|
||||
"""HTTP helper that checks the invariants every response must satisfy."""
|
||||
|
||||
def __init__(self, base_url: str = BASE_URL):
|
||||
self.client = httpx.Client(base_url=base_url, timeout=15)
|
||||
|
||||
def call(self, method: str, path: str, expected: int | tuple[int, ...], *, token: str | None = None,
|
||||
json_body=None, raw: bytes | None = None, headers: dict | None = None, schema: str | None = None,
|
||||
lb_error_ok: bool = False):
|
||||
"""lb_error_ok: a 502/503/504 generated by HAProxy itself (no API replica answered) is
|
||||
accepted without X-Instance-Id; every response of an API replica must carry it."""
|
||||
h = dict(headers or {})
|
||||
if token is not None:
|
||||
h["Authorization"] = "Bearer " + token
|
||||
if json_body is not None:
|
||||
raw = json.dumps(json_body, ensure_ascii=False).encode("utf-8")
|
||||
h.setdefault("Content-Type", "application/json")
|
||||
response = self.client.request(method, path, content=raw, headers=h)
|
||||
allowed = (expected,) if isinstance(expected, int) else expected
|
||||
assert response.status_code in allowed, (method, path, response.status_code, response.text[:300])
|
||||
request_id = response.headers.get("x-request-id")
|
||||
assert request_id, "every response must carry X-Request-Id"
|
||||
if not response.headers.get("x-instance-id"):
|
||||
assert lb_error_ok and response.status_code in (502, 503, 504), \
|
||||
("missing X-Instance-Id", method, path, response.status_code)
|
||||
if response.status_code >= 400:
|
||||
body = response.json()
|
||||
assert schema_errors(body, "Error") == [], response.text
|
||||
assert body["error"]["request_id"] == request_id
|
||||
assert "Traceback" not in response.text and "psycopg" not in response.text
|
||||
if response.status_code == 401:
|
||||
challenge = "basic" if path == "/api/v1/auth/sessions" else "bearer"
|
||||
assert response.headers.get("www-authenticate", "").lower().startswith(challenge)
|
||||
elif schema:
|
||||
assert schema_errors(response.json(), schema) == [], response.text
|
||||
return response
|
||||
|
||||
def register(self, label: str = "u", *, password: str = PASSWORD, display_name: str | None = None) -> dict:
|
||||
username = f"{label}_{uuid.uuid4().hex[:12]}"[:32]
|
||||
body = {"username": username, "display_name": display_name or f"Студент {label}", "password": password}
|
||||
user = self.call("POST", "/api/v1/users", 201, json_body=body, schema="User").json()
|
||||
return {"id": user["id"], "user": user, "username": username, "password": password}
|
||||
|
||||
def login(self, actor: dict) -> str:
|
||||
session = self.call("POST", "/api/v1/auth/sessions", 201,
|
||||
headers={"Authorization": basic(actor["username"], actor["password"])},
|
||||
schema="Session").json()
|
||||
assert session["user"] == actor["user"] and session["token_type"] == "Bearer"
|
||||
return session["token"]
|
||||
|
||||
def user(self, label: str = "u", **kwargs) -> dict:
|
||||
actor = self.register(label, **kwargs)
|
||||
actor["token"] = self.login(actor)
|
||||
return actor
|
||||
|
||||
def chat(self, owner: dict, *members: dict, title: str = "Тестовый чат") -> dict:
|
||||
ids = [owner["id"], *(m["id"] for m in members)]
|
||||
return self.call("POST", "/api/v1/chats", 201, token=owner["token"],
|
||||
json_body={"title": title, "member_ids": ids}, schema="Chat").json()
|
||||
|
||||
def post_message(self, actor: dict, chat: dict, body, expected=201, *, key: str | None = None,
|
||||
raw: bytes | None = None, headers: dict | None = None, **kwargs):
|
||||
"""POST a message with an Idempotency-Key (a fresh one unless `key` is given)."""
|
||||
h = {"Idempotency-Key": key or uuid.uuid4().hex, **(headers or {})}
|
||||
if raw is not None:
|
||||
h.setdefault("Content-Type", "application/json")
|
||||
return self.call("POST", f"/api/v1/chats/{chat['id']}/messages", expected, token=actor["token"],
|
||||
json_body=body if raw is None else None, raw=raw, headers=h,
|
||||
schema="Message" if expected == 201 else None, **kwargs)
|
||||
|
||||
def send(self, actor: dict, chat: dict, text: str, key: str | None = None) -> dict:
|
||||
return self.post_message(actor, chat, {"text": text}, key=key).json()
|
||||
|
||||
def history(self, actor: dict, chat: dict, limit: int) -> tuple[list[dict], int]:
|
||||
"""Walk all pages; returns (items, number_of_pages)."""
|
||||
items, pages, cursor = [], 0, None
|
||||
while True:
|
||||
query = f"?limit={limit}" + (f"&cursor={cursor}" if cursor else "")
|
||||
page = self.call("GET", f"/api/v1/chats/{chat['id']}/messages{query}", 200,
|
||||
token=actor["token"], schema="MessagePage").json()
|
||||
pages += 1
|
||||
items.extend(page["items"])
|
||||
cursor = page["next_cursor"]
|
||||
if cursor is None:
|
||||
return items, pages
|
||||
assert len(page["items"]) == limit, "a non-final page must be full"
|
||||
assert len(items) < 100_000, "cursor loop"
|
||||
|
||||
|
||||
def _stand_is_up() -> bool:
|
||||
try:
|
||||
return httpx.get(BASE_URL + "/health/ready", timeout=3).status_code == 200
|
||||
except httpx.HTTPError:
|
||||
return False
|
||||
|
||||
|
||||
def pytest_collection_modifyitems(config, items):
|
||||
if any("stack" in item.keywords for item in items) and not _stand_is_up():
|
||||
skip = pytest.mark.skip(reason=f"stand is not running at {BASE_URL}: docker compose up -d --wait")
|
||||
for item in items:
|
||||
if "stack" in item.keywords:
|
||||
item.add_marker(skip)
|
||||
|
||||
|
||||
@pytest.fixture(scope="session")
|
||||
def api() -> Api:
|
||||
return Api()
|
||||
|
||||
|
||||
def run_hook(name: str, timeout: float = 180, **env: str) -> subprocess.CompletedProcess:
|
||||
"""Run scripts/<name> (a POSIX sh hook); on Windows through Git Bash like the .cmd wrappers."""
|
||||
script = str(ROOT / "scripts" / name)
|
||||
if os.name == "nt":
|
||||
command = [os.path.join(os.environ.get("ProgramFiles", r"C:\Program Files"), "Git", "bin", "bash.exe"), script]
|
||||
else:
|
||||
command = [script]
|
||||
return subprocess.run(command, cwd=ROOT, capture_output=True, text=True, encoding="utf-8", check=True,
|
||||
timeout=timeout, env={**os.environ, **env})
|
||||
|
||||
|
||||
def probe_inside(replica: dict, path: str) -> int:
|
||||
"""HTTP status of `path` asked inside the replica container itself (bypassing the LB)."""
|
||||
code = ("import urllib.request as u, urllib.error as e\n"
|
||||
"try: print(u.urlopen('http://127.0.0.1:8080%s', timeout=3).status)\n"
|
||||
"except e.HTTPError as x: print(x.code)") % path
|
||||
out = subprocess.run(["docker", "exec", replica["id"], "python", "-c", code], capture_output=True, text=True,
|
||||
timeout=30).stdout.strip()
|
||||
return int(out or 0)
|
||||
@@ -0,0 +1,344 @@
|
||||
"""Contract behaviour over real HTTP (stand must be running). Lab 1 checks moved to Bearer + new lab 2 auth."""
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import uuid
|
||||
|
||||
import pytest
|
||||
|
||||
from conftest import PASSWORD, basic
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
|
||||
def key(item: dict) -> tuple[str, int]:
|
||||
return item["created_at"], uuid.UUID(item["id"]).int
|
||||
|
||||
|
||||
# ================================================================== registration & login
|
||||
def test_register_never_returns_password_and_login_returns_bearer(api):
|
||||
actor = api.register("alice", display_name="Алиса 👩💻")
|
||||
assert set(actor["user"]) == {"id", "username", "display_name", "created_at"}
|
||||
session = api.call("POST", "/api/v1/auth/sessions", 201,
|
||||
headers={"Authorization": basic(actor["username"], PASSWORD)}, schema="Session").json()
|
||||
assert session["token_type"] == "Bearer" and len(session["token"]) >= 43
|
||||
assert actor["username"] not in session["token"] and actor["id"] not in session["token"]
|
||||
assert api.call("GET", "/api/v1/users/me", 200, token=session["token"], schema="User").json() == actor["user"]
|
||||
|
||||
|
||||
def test_duplicate_username_is_409(api):
|
||||
actor = api.register("dup")
|
||||
body = {"username": actor["username"], "display_name": "Другой", "password": PASSWORD}
|
||||
assert api.call("POST", "/api/v1/users", 409, json_body=body).json()["error"]["code"] == "conflict"
|
||||
|
||||
|
||||
@pytest.mark.parametrize("password", ["short-11ch", " " * 12, " \t \n " * 3, "я" * 129])
|
||||
def test_bad_password_on_registration_is_400(api, password):
|
||||
body = {"username": "pw_" + uuid.uuid4().hex[:10], "display_name": "x", "password": password}
|
||||
api.call("POST", "/api/v1/users", 400, json_body=body)
|
||||
|
||||
|
||||
@pytest.mark.parametrize("password", ["я" * 128, "with:colons:inside", "Twelve chars"])
|
||||
def test_edge_passwords_register_and_login(api, password):
|
||||
actor = api.user("edge", password=password)
|
||||
api.call("GET", "/api/v1/users/me", 200, token=actor["token"])
|
||||
|
||||
|
||||
def test_wrong_password_and_unknown_user_give_the_same_401(api):
|
||||
actor = api.register("victim")
|
||||
wrong = api.call("POST", "/api/v1/auth/sessions", 401, headers={"Authorization": basic(actor["username"], "Wrong-password-1")})
|
||||
unknown = api.call("POST", "/api/v1/auth/sessions", 401, headers={"Authorization": basic("nobody_" + uuid.uuid4().hex[:8], PASSWORD)})
|
||||
assert wrong.json()["error"]["message"] == unknown.json()["error"]["message"]
|
||||
assert wrong.headers["www-authenticate"] == 'Basic realm="messenger", charset="UTF-8"'
|
||||
|
||||
|
||||
@pytest.mark.parametrize("header", ["Basic é", "Basic !!!", "Basic " + base64.b64encode(b"no_colon").decode(),
|
||||
"Basic " + base64.b64encode(b"\xff\xfe:x").decode(), "Bearer " + "a" * 43, ""])
|
||||
def test_malformed_basic_is_401_not_500(api, header):
|
||||
# Raw UTF-8 bytes, as curl would send them (httpx refuses non-ASCII str headers).
|
||||
api.call("POST", "/api/v1/auth/sessions", 401, headers={"Authorization": header.encode("utf-8")} if header else {})
|
||||
|
||||
|
||||
def test_logout_revokes_only_this_token(api):
|
||||
actor = api.user("logout")
|
||||
second = api.login(actor)
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 204, token=actor["token"])
|
||||
api.call("GET", "/api/v1/users/me", 401, token=actor["token"])
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 401, token=actor["token"])
|
||||
api.call("GET", "/api/v1/users/me", 200, token=second) # other session of the same user survives
|
||||
|
||||
|
||||
def test_x_user_id_basic_and_garbage_do_not_authenticate(api):
|
||||
actor = api.user("hdr")
|
||||
for headers in ({"X-User-Id": actor["id"]},
|
||||
{"Authorization": basic(actor["username"], PASSWORD)},
|
||||
{"Authorization": "Bearer " + "A" * 43},
|
||||
{"Authorization": "Bearer"}):
|
||||
api.call("GET", "/api/v1/users/me", 401, headers=headers)
|
||||
me = api.call("GET", "/api/v1/users/me", 200, token=actor["token"], headers={"X-User-Id": str(uuid.uuid4())})
|
||||
assert me.json()["id"] == actor["id"]
|
||||
|
||||
|
||||
# ================================================================== validation / errors (from lab 1)
|
||||
@pytest.mark.parametrize("raw", [b"{", b"", b"[]", b"null", b'{"username": "abc_def", "display_name": NaN, "password": "Twelve-chars"}',
|
||||
b'{"username": "abc_def", "display_name": "x", "password": "Twelve-chars"}{"a": 1}',
|
||||
b'{"username": "abc_def", "display_name": "\\ud800", "password": "Twelve-chars"}'])
|
||||
def test_malformed_json_is_400(api, raw):
|
||||
api.call("POST", "/api/v1/users", 400, raw=raw, headers={"Content-Type": "application/json"})
|
||||
|
||||
|
||||
@pytest.mark.parametrize("body", [
|
||||
{"username": "abc_def", "display_name": "x"}, # no password
|
||||
{"username": "abc_def", "display_name": "x", "password": "Twelve-chars", "extra": 1},
|
||||
{"username": "Abc_def", "display_name": "x", "password": "Twelve-chars"},
|
||||
{"username": "abc_def", "display_name": " ", "password": "Twelve-chars"},
|
||||
{"username": "abc_def", "display_name": "x" * 101, "password": "Twelve-chars"},
|
||||
{"username": 5, "display_name": "x", "password": "Twelve-chars"},
|
||||
])
|
||||
def test_invalid_registration_fields_are_400(api, body):
|
||||
api.call("POST", "/api/v1/users", 400, json_body=body)
|
||||
|
||||
|
||||
def test_unicode_limits_are_code_points_and_nul_is_stored(api):
|
||||
actor = api.user("uni", display_name="😀" * 100)
|
||||
assert actor["user"]["display_name"] == "😀" * 100
|
||||
nul = api.user("nul", display_name="A\u0000B")
|
||||
assert api.call("GET", "/api/v1/users/me", 200, token=nul["token"]).json()["display_name"] == "A\u0000B"
|
||||
chat = api.chat(actor, title="T\u0000itle")
|
||||
assert chat["title"] == "T\u0000itle"
|
||||
assert api.send(actor, chat, "hello\u0000world")["text"] == "hello\u0000world"
|
||||
assert api.send(actor, chat, "🙂" * 4000)["text"] == "🙂" * 4000
|
||||
api.post_message(actor, chat, {"text": "я" * 4001}, 400)
|
||||
texts = [m["text"] for m in api.history(actor, chat, 100)[0]]
|
||||
assert texts == ["hello\u0000world", "🙂" * 4000]
|
||||
|
||||
|
||||
def test_oversized_body_and_wrong_content_type(api):
|
||||
api.call("POST", "/api/v1/users", 413, json_body={"username": "big_body", "display_name": "x" * 70_000, "password": PASSWORD})
|
||||
api.call("POST", "/api/v1/users", 400, raw=b'{"username":"abc_def"}', headers={"Content-Type": "text/plain"})
|
||||
|
||||
|
||||
# ================================================================== chats & membership
|
||||
def test_chat_validation(api):
|
||||
alice, bob = api.user("alice"), api.user("bob")
|
||||
for body in ({"title": "x", "member_ids": [alice["id"], bob["id"], bob["id"]]},
|
||||
{"title": "x", "member_ids": [alice["id"], str(uuid.uuid4())]},
|
||||
{"title": "x", "member_ids": [bob["id"]]},
|
||||
{"title": " ", "member_ids": [alice["id"]]},
|
||||
{"title": "x", "member_ids": [alice["id"], "bob"]}):
|
||||
api.call("POST", "/api/v1/chats", 400, token=alice["token"], json_body=body)
|
||||
|
||||
|
||||
def test_foreign_chat_is_indistinguishable_from_missing(api):
|
||||
alice, bob, eve = api.user("alice"), api.user("bob"), api.user("eve")
|
||||
chat = api.chat(alice, bob)
|
||||
for chat_id in (chat["id"], str(uuid.uuid4())):
|
||||
base = f"/api/v1/chats/{chat_id}"
|
||||
errors = [api.call("GET", base, 404, token=eve["token"]).json(),
|
||||
api.call("GET", base + "/messages", 404, token=eve["token"]).json(),
|
||||
api.call("POST", base + "/messages", 404, token=eve["token"], json_body={"text": "чужое"}, headers={"Idempotency-Key": "k1"}).json(),
|
||||
api.call("GET", base + "/messages?cursor=garbage", 404, token=eve["token"]).json()]
|
||||
assert len({e["error"]["message"] for e in errors}) == 1
|
||||
api.call("GET", "/api/v1/chats/not-a-uuid", 401)
|
||||
api.call("GET", "/api/v1/chats/not-a-uuid", 400, token=alice["token"])
|
||||
api.call("GET", "/api/v1/chats/NOT-A-UUID/messages", 400, token=alice["token"])
|
||||
|
||||
|
||||
def test_chat_list_order_pagination_and_isolation(api):
|
||||
alice, bob, eve = api.user("alice"), api.user("bob"), api.user("eve")
|
||||
created = [api.chat(alice, bob, title=f"Чат {i}") for i in range(5)]
|
||||
api.chat(eve, title="Чат Евы")
|
||||
walked, cursor = [], None
|
||||
while True:
|
||||
query = "?limit=2" + (f"&cursor={cursor}" if cursor else "")
|
||||
page = api.call("GET", "/api/v1/chats" + query, 200, token=bob["token"], schema="ChatPage").json()
|
||||
walked += page["items"]
|
||||
cursor = page["next_cursor"]
|
||||
if not cursor:
|
||||
break
|
||||
assert walked == created
|
||||
first = api.call("GET", "/api/v1/chats?limit=1", 200, token=alice["token"]).json()["next_cursor"]
|
||||
api.call("GET", f"/api/v1/chats?limit=1&cursor={first}", 400, token=bob["token"]) # cursor of another user
|
||||
api.call("GET", "/api/v1/chats?limit=0", 400, token=bob["token"])
|
||||
|
||||
|
||||
# ================================================================== messages & keyset pagination
|
||||
def test_history_pages_without_loss_or_duplicates(api):
|
||||
alice, bob = api.user("alice"), api.user("bob")
|
||||
chat = api.chat(alice, bob)
|
||||
sent = [api.send(alice if i % 2 else bob, chat, f"сообщение {i}") for i in range(25)]
|
||||
items, pages = api.history(bob, chat, 7)
|
||||
assert pages == 4 and items == sent and items == sorted(items, key=key)
|
||||
assert sent[0]["sender_id"] == bob["id"] and sent[1]["sender_id"] == alice["id"]
|
||||
|
||||
|
||||
@pytest.mark.parametrize("query", ["limit=0", "limit=101", "limit=abc", "limit=1&limit=2", "cursor=", "cursor=garbage",
|
||||
"cursor=djF8bXx4fDF8Mg", "cursor=" + base64.urlsafe_b64encode(b"v1|m|x|99999999999999999999|y").decode()])
|
||||
def test_bad_limit_or_cursor_is_400(api, query):
|
||||
alice = api.user("alice")
|
||||
chat = api.chat(alice)
|
||||
api.call("GET", f"/api/v1/chats/{chat['id']}/messages?{query}", 400, token=alice["token"])
|
||||
|
||||
|
||||
def test_cursor_bound_to_chat(api):
|
||||
alice = api.user("alice")
|
||||
a, b = api.chat(alice, title="A"), api.chat(alice, title="B")
|
||||
for chat in (a, b):
|
||||
api.send(alice, chat, "1")
|
||||
api.send(alice, chat, "2")
|
||||
cursor = api.call("GET", f"/api/v1/chats/{a['id']}/messages?limit=1", 200, token=alice["token"]).json()["next_cursor"]
|
||||
api.call("GET", f"/api/v1/chats/{b['id']}/messages?cursor={cursor}", 400, token=alice["token"])
|
||||
empty = api.chat(alice, title="Empty")
|
||||
api.call("GET", f"/api/v1/chats/{empty['id']}/messages?cursor=garbage", 400, token=alice["token"])
|
||||
|
||||
|
||||
def test_read_after_write(api):
|
||||
alice, bob = api.user("alice"), api.user("bob")
|
||||
chat = api.chat(alice, bob)
|
||||
for i in range(5):
|
||||
message = api.send(alice, chat, f"m{i}")
|
||||
last = api.call("GET", f"/api/v1/chats/{chat['id']}/messages?limit=100", 200, token=bob["token"]).json()["items"][-1]
|
||||
assert last == message
|
||||
|
||||
|
||||
# ================================================================== lab 3: inputs found by the reviewer in other PRs
|
||||
def b64(raw: bytes) -> str:
|
||||
return base64.urlsafe_b64encode(raw).rstrip(b"=").decode()
|
||||
|
||||
|
||||
def bad_cursors(owner_id: str) -> list[str]:
|
||||
"""Wrong types/values inside the cursor, empty and whitespace cursors, out-of-range time."""
|
||||
good_uuid = str(uuid.uuid4())
|
||||
return ["", " ", "%20", "garbage", b64(b"[]"), b64(b"{}"), b64(b"123"), b64(b"false"), b64(b"null"),
|
||||
b64(b'{"created_at":1,"id":1}'), b64(b'{"created_at":null,"id":"x"}'),
|
||||
b64(f"v1|m|{owner_id}|1|1".encode()), b64(f"v1|c|{owner_id}|1|1".encode()),
|
||||
b64(f"v1|m|{owner_id}|abc|{good_uuid}".encode()), b64(f"v1|m|{owner_id}|-1|{good_uuid}".encode()),
|
||||
b64(f"v1|m|{owner_id}|9223372036854775807|{good_uuid}".encode()),
|
||||
b64(f"v1|c|{owner_id}|9223372036854775807|{good_uuid}".encode()),
|
||||
b64(f"v1|m|{owner_id}|1|{good_uuid.upper()}".encode()), b64(f"v2|m|{owner_id}|1|{good_uuid}".encode())]
|
||||
|
||||
|
||||
def test_malformed_cursor_values_are_400_on_both_lists(api):
|
||||
alice = api.user("cur")
|
||||
chat = api.chat(alice)
|
||||
for cursor in bad_cursors(chat["id"]) + bad_cursors(alice["id"]):
|
||||
quoted = cursor if cursor in ("", "%20") else cursor.replace(" ", "%20")
|
||||
for path in (f"/api/v1/chats/{chat['id']}/messages", "/api/v1/chats"):
|
||||
error = api.call("GET", f"{path}?cursor={quoted}", 400, token=alice["token"]).json()
|
||||
assert error["error"]["code"] == "invalid_request", (path, cursor)
|
||||
|
||||
|
||||
@pytest.mark.parametrize("raw", [b"\xff", b'{"text":"first","TEXT":"second"}', b'{"text":"a","text":"b"}',
|
||||
b'{"text":123}', b'{"text":"valid"}}', b'{"text":"valid"}]', b'{"text":"x"} {}',
|
||||
b'{"text":"x","sender_id":"11111111-1111-4111-8111-111111111111"}',
|
||||
b'{"text":"x","attachment_ids":null}', b'{"text":"x","attachment_ids":["nope"]}',
|
||||
b'{"text":" \n\t "}'])
|
||||
def test_bad_message_body_is_400_and_nothing_is_stored(api, raw):
|
||||
alice = api.user("body")
|
||||
chat = api.chat(alice)
|
||||
error = api.post_message(alice, chat, None, 400, raw=raw).json()
|
||||
assert error["error"]["code"] == "invalid_request"
|
||||
assert api.history(alice, chat, 100)[0] == []
|
||||
|
||||
|
||||
def test_ff_byte_body_on_registration_is_contract_error(api):
|
||||
response = api.call("POST", "/api/v1/users", 400, raw=b"\xff", headers={"Content-Type": "application/json"})
|
||||
assert response.json()["error"]["request_id"] == response.headers["x-request-id"]
|
||||
|
||||
|
||||
def test_text_is_stored_verbatim_and_length_counts_raw_text(api):
|
||||
alice = api.user("verbatim")
|
||||
chat = api.chat(alice)
|
||||
padded = " строка\nс пробелами "
|
||||
assert api.send(alice, chat, padded)["text"] == padded
|
||||
api.post_message(alice, chat, {"text": " " + "a" * 4000}, 400) # 4001 code points before any trimming
|
||||
assert [m["text"] for m in api.history(alice, chat, 100)[0]] == [padded]
|
||||
|
||||
|
||||
def test_unknown_attachment_ids_are_404(api):
|
||||
alice = api.user("att")
|
||||
chat = api.chat(alice)
|
||||
api.post_message(alice, chat, {"text": "x", "attachment_ids": [str(uuid.uuid4())]}, 404)
|
||||
api.post_message(alice, chat, {"text": "x", "attachment_ids": [str(uuid.uuid4())] * 2}, 400) # duplicates
|
||||
assert api.history(alice, chat, 100)[0] == []
|
||||
|
||||
|
||||
def test_member_ids_are_compared_as_uuids_not_strings(api):
|
||||
alice, bob = api.user("case"), api.user("case")
|
||||
api.call("POST", "/api/v1/chats", 400, token=alice["token"],
|
||||
json_body={"title": "x", "member_ids": [alice["id"], alice["id"].upper()]})
|
||||
api.call("POST", "/api/v1/chats", 400, token=alice["token"],
|
||||
json_body={"title": "x", "member_ids": [alice["id"], bob["id"]], "id": str(uuid.uuid4())})
|
||||
chat = api.call("POST", "/api/v1/chats", 201, token=alice["token"],
|
||||
json_body={"title": "x", "member_ids": [alice["id"].upper(), bob["id"]]}, schema="Chat").json()
|
||||
assert chat["member_ids"] == [alice["id"], bob["id"]]
|
||||
|
||||
|
||||
@pytest.mark.parametrize("header", ["Bearer é", "Bearer .", "Bearer a b", "Bearer " + "é" * 43, "bearer\t" + "a" * 43])
|
||||
def test_non_ascii_or_garbage_bearer_is_401_not_500(api, header):
|
||||
api.call("GET", "/api/v1/users/me", 401, headers={"Authorization": header.encode("utf-8")})
|
||||
|
||||
|
||||
def test_login_with_a_body_is_rejected(api):
|
||||
actor = api.register("login_body")
|
||||
auth = {"Authorization": basic(actor["username"], actor["password"])}
|
||||
api.call("POST", "/api/v1/auth/sessions", 400, raw=b'{"extra":1}', headers={**auth, "Content-Type": "application/json"})
|
||||
api.call("POST", "/api/v1/auth/sessions", 400, raw=b"x", headers=auth)
|
||||
api.call("POST", "/api/v1/auth/sessions", 201, headers=auth, schema="Session")
|
||||
|
||||
|
||||
def test_password_range_is_12_to_128_code_points(api):
|
||||
api.call("POST", "/api/v1/users", 400, json_body={"username": "pw_" + uuid.uuid4().hex[:10], "display_name": "x", "password": "a"})
|
||||
actor = api.user("pw128", password="я:" * 64)
|
||||
api.call("GET", "/api/v1/users/me", 200, token=actor["token"])
|
||||
|
||||
|
||||
def test_every_api_response_carries_instance_and_request_id(api):
|
||||
"""X-Instance-Id also on 404 of unknown routes, 405, 401, 400 and health (Api.call asserts it)."""
|
||||
api.call("GET", "/health/live", 200)
|
||||
api.call("GET", "/health/ready", 200)
|
||||
api.call("GET", "/no/such/route", 404)
|
||||
api.call("PUT", "/api/v1/users", 405)
|
||||
api.call("GET", "/api/v1/users/me", 401)
|
||||
response = api.call("GET", "/api/v1/users/me", 401, headers={"X-Request-Id": "client-supplied.id-1"})
|
||||
assert response.headers["x-request-id"] == "client-supplied.id-1" # one id from LB log to API log
|
||||
|
||||
|
||||
@pytest.mark.parametrize("blank", [" ", " ", " \t "])
|
||||
def test_unicode_whitespace_only_strings_are_400(api, blank):
|
||||
alice = api.user("blank")
|
||||
chat = api.chat(alice)
|
||||
api.call("POST", "/api/v1/users", 400, json_body={"username": "bl_" + uuid.uuid4().hex[:10], "display_name": blank,
|
||||
"password": PASSWORD})
|
||||
api.call("POST", "/api/v1/chats", 400, token=alice["token"], json_body={"title": blank, "member_ids": [alice["id"]]})
|
||||
api.post_message(alice, chat, {"text": blank}, 400)
|
||||
assert api.history(alice, chat, 100)[0] == []
|
||||
|
||||
|
||||
@pytest.mark.parametrize("raw", [b'{"username":"num_probe","display_name":"x","password":1e1000}',
|
||||
'{"username":"ff_probe","display_name":"x","password":"12345678901'.encode() + b'\xff"}',
|
||||
b'{"username":"num_probe","display_name":1,"password":"Twelve-chars-ok"}'])
|
||||
def test_odd_registration_bodies_are_400_and_the_api_survives(api, raw):
|
||||
api.call("POST", "/api/v1/users", 400, raw=raw, headers={"Content-Type": "application/json"})
|
||||
api.call("GET", "/health/ready", 200)
|
||||
|
||||
|
||||
def test_more_than_100_members_is_400(api):
|
||||
owner = api.user("big")
|
||||
others = [api.register(f"m{i}")["id"] for i in range(100)]
|
||||
api.call("POST", "/api/v1/chats", 400, token=owner["token"], json_body={"title": "x", "member_ids": [owner["id"], *others]})
|
||||
chat = api.call("POST", "/api/v1/chats", 201, token=owner["token"],
|
||||
json_body={"title": "x", "member_ids": [owner["id"], *others[:99]]}, schema="Chat").json()
|
||||
assert len(chat["member_ids"]) == 100
|
||||
|
||||
|
||||
def test_cursor_of_one_list_is_rejected_by_the_other(api):
|
||||
alice = api.user("cross")
|
||||
chats = [api.chat(alice, title=f"c{i}") for i in range(2)]
|
||||
for _ in range(2):
|
||||
api.send(alice, chats[0], "m")
|
||||
message_cursor = api.call("GET", f"/api/v1/chats/{chats[0]['id']}/messages?limit=1", 200,
|
||||
token=alice["token"]).json()["next_cursor"]
|
||||
chat_cursor = api.call("GET", "/api/v1/chats?limit=1", 200, token=alice["token"]).json()["next_cursor"]
|
||||
api.call("GET", f"/api/v1/chats?cursor={message_cursor}", 400, token=alice["token"])
|
||||
api.call("GET", f"/api/v1/chats/{chats[0]['id']}/messages?cursor={chat_cursor}", 400, token=alice["token"])
|
||||
@@ -0,0 +1,325 @@
|
||||
"""Load balancer, replicas and graceful shutdown (lab 3, levels 3 and 4).
|
||||
|
||||
Each test that stops, kills, rescales or isolates a replica restores the stand in `finally`
|
||||
and waits until HAProxy reports every replica UP again (conftest.wait_stand).
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import contextlib
|
||||
import json
|
||||
import subprocess
|
||||
import threading
|
||||
import time
|
||||
import uuid
|
||||
|
||||
import httpx2 as httpx
|
||||
import pytest
|
||||
|
||||
from conftest import BASE_URL, REPLICAS, Api, compose, lbctl, probe_inside, psql, replica_stopped, run_hook, \
|
||||
running_replicas, wait_healthy, wait_stand
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
|
||||
def instances_seen(api: Api, token: str, requests: int = 40) -> list[str]:
|
||||
return [api.call("GET", "/api/v1/users/me", 200, token=token).headers["x-instance-id"] for _ in range(requests)]
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ level 3: replicas, no affinity
|
||||
def test_one_session_is_served_by_every_real_replica(api):
|
||||
actor = api.user("rr")
|
||||
seen = instances_seen(api, actor["token"])
|
||||
running = {r["hostname"] for r in running_replicas()}
|
||||
assert len(running) == REPLICAS >= 2
|
||||
assert set(seen) == running, "X-Instance-Id values must be exactly the hostnames of the running containers"
|
||||
# Round robin on ONE keep-alive client connection: consecutive requests switch replicas,
|
||||
# so there is no sticky session hiding split state.
|
||||
switches = sum(1 for a, b in zip(seen, seen[1:]) if a != b)
|
||||
assert switches >= len(seen) // 2, seen
|
||||
assert {s["instance_id"] for s in lbctl.lb_servers() if s["state"] == "UP"} == running
|
||||
|
||||
|
||||
def test_write_on_one_replica_is_read_on_the_other(api):
|
||||
alice, bob = api.user("rw"), api.user("rw")
|
||||
chat = api.chat(alice, bob)
|
||||
for i in range(10):
|
||||
sent = api.post_message(alice, chat, {"text": f"m{i}"})
|
||||
read = api.call("GET", f"/api/v1/chats/{chat['id']}/messages?limit=100", 200, token=bob["token"])
|
||||
assert read.json()["items"][-1] == sent.json()
|
||||
logout = api.call("DELETE", "/api/v1/auth/sessions/current", 204, token=alice["token"])
|
||||
for _ in range(6): # the revocation is visible on every replica at once (shared Valkey)
|
||||
r = api.call("GET", "/api/v1/users/me", 401, token=alice["token"])
|
||||
assert r.headers["x-instance-id"]
|
||||
assert logout.headers["x-instance-id"]
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ readiness drives the LB
|
||||
def hba_file() -> str:
|
||||
return psql("SHOW hba_file")
|
||||
|
||||
|
||||
@contextlib.contextmanager
|
||||
def isolated_from_postgres(replica: dict):
|
||||
"""Make ONE replica unable to reach PostgreSQL (pg_hba reject for its IP + terminate its
|
||||
connections), while it keeps running and answering /health/live. Always undone."""
|
||||
ip = replica["networks"]["data"] # PostgreSQL sees the replica's address on the data network
|
||||
path = hba_file()
|
||||
marker = "lab3-test-isolate"
|
||||
compose("exec", "-T", "postgres", "sh", "-c", f"sed -i '1i host all all {ip}/32 reject # {marker}' {path}")
|
||||
try:
|
||||
psql("SELECT pg_reload_conf()")
|
||||
psql(f"SELECT count(pg_terminate_backend(pid)) FROM pg_stat_activity WHERE client_addr = '{ip}'")
|
||||
yield replica
|
||||
finally:
|
||||
compose("exec", "-T", "postgres", "sh", "-c", f"sed -i '/{marker}/d' {path}")
|
||||
psql("SELECT pg_reload_conf()")
|
||||
lbctl.wait_up(REPLICAS, timeout=15) # self-heals: no restart of the replica needed
|
||||
wait_stand()
|
||||
|
||||
|
||||
@contextlib.contextmanager
|
||||
def disconnected_from_data_network(replica: dict):
|
||||
"""`docker network disconnect <project>_data <api>`: the replica keeps its edge link to the
|
||||
LB but loses PostgreSQL AND Valkey. Always reconnected."""
|
||||
network = next(n for n in json.loads(subprocess.run(
|
||||
["docker", "inspect", "-f", "{{json .NetworkSettings.Networks}}", replica["id"]],
|
||||
capture_output=True, text=True, check=True).stdout) if n.endswith("_data"))
|
||||
subprocess.run(["docker", "network", "disconnect", network, replica["id"]], check=True, capture_output=True)
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
subprocess.run(["docker", "network", "connect", network, replica["id"]], check=True, capture_output=True)
|
||||
lbctl.wait_up(REPLICAS, timeout=15) # self-heals: dead sockets are dropped (tcp_user_timeout)
|
||||
wait_stand()
|
||||
|
||||
|
||||
@pytest.mark.parametrize("how", ["pg_hba", "network"])
|
||||
@pytest.mark.parametrize("position", range(REPLICAS))
|
||||
def test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready(api, position, how):
|
||||
"""The reviewers' scenarios: one API loses PostgreSQL (pg_hba reject) or its whole data
|
||||
network (docker network disconnect) but keeps running. /health/live 200, /health/ready
|
||||
503 -> HAProxy must take it out; no business request may reach it."""
|
||||
replica = sorted(running_replicas(), key=lambda r: r["name"])[position]
|
||||
actor = api.user("ready")
|
||||
isolate = isolated_from_postgres if how == "pg_hba" else disconnected_from_data_network
|
||||
with isolate(replica):
|
||||
started = time.monotonic()
|
||||
while replica["hostname"] in lbctl.up_instances():
|
||||
assert time.monotonic() - started < 5, "HAProxy did not mark the unready replica DOWN within 5 s"
|
||||
time.sleep(0.1)
|
||||
detected = time.monotonic() - started
|
||||
for _ in range(3): # the process stays responsive (event loop never blocked) ...
|
||||
started_probe = time.monotonic()
|
||||
assert probe_inside(replica, "/health/live") == 200
|
||||
assert probe_inside(replica, "/health/ready") == 503 # ... but not ready
|
||||
assert time.monotonic() - started_probe < 4.5, "replica froze while its dependencies were unreachable"
|
||||
seen = instances_seen(api, actor["token"], 40)
|
||||
assert replica["hostname"] not in seen and len(seen) == 40
|
||||
for i in range(20): # writes keep working, only through the ready replicas
|
||||
assert api.register(f"r{i}")["id"]
|
||||
print(f"{replica['name']} ({how}): unready replica out of rotation after {detected:.2f} s")
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ stopping replicas under traffic
|
||||
class Traffic:
|
||||
"""Background writers/readers through the LB; uncertain POSTs are retried with the SAME key."""
|
||||
|
||||
def __init__(self, api: Api, workers: int = 4):
|
||||
self.alice, self.bob = api.user("load"), api.user("load")
|
||||
self.chat = api.chat(self.alice, self.bob)
|
||||
self.workers = workers
|
||||
self.stop = threading.Event()
|
||||
self.lock = threading.Lock()
|
||||
self.events: list[tuple[float, str, int, str | None]] = [] # (time, kind, status, instance)
|
||||
self.confirmed: list[dict] = []
|
||||
self.threads = [threading.Thread(target=self.run, args=(i,), daemon=True) for i in range(workers)]
|
||||
|
||||
def record(self, kind: str, status: int, instance: str | None) -> None:
|
||||
with self.lock:
|
||||
self.events.append((time.monotonic(), kind, status, instance))
|
||||
|
||||
def run(self, index: int) -> None:
|
||||
client = httpx.Client(base_url=BASE_URL, timeout=3)
|
||||
auth = {"Authorization": "Bearer " + self.alice["token"]}
|
||||
n = 0
|
||||
key, text = uuid.uuid4().hex, f"w{index}-0"
|
||||
while not self.stop.is_set():
|
||||
try:
|
||||
r = client.post(f"/api/v1/chats/{self.chat['id']}/messages", json={"text": text},
|
||||
headers={**auth, "Idempotency-Key": key})
|
||||
self.record("POST", r.status_code, r.headers.get("x-instance-id"))
|
||||
if r.status_code == 201:
|
||||
with self.lock:
|
||||
self.confirmed.append(r.json())
|
||||
n += 1
|
||||
key, text = uuid.uuid4().hex, f"w{index}-{n}"
|
||||
# any other outcome: keep key and text, retry the same request
|
||||
except httpx.HTTPError:
|
||||
self.record("POST", 0, None)
|
||||
try:
|
||||
r = client.get(f"/api/v1/chats/{self.chat['id']}/messages?limit=5", headers=auth)
|
||||
self.record("GET", r.status_code, r.headers.get("x-instance-id"))
|
||||
except httpx.HTTPError:
|
||||
self.record("GET", 0, None)
|
||||
time.sleep(0.02)
|
||||
client.close()
|
||||
|
||||
def __enter__(self) -> "Traffic":
|
||||
for t in self.threads:
|
||||
t.start()
|
||||
time.sleep(1.0)
|
||||
return self
|
||||
|
||||
def __exit__(self, *exc) -> None:
|
||||
self.stop.set()
|
||||
for t in self.threads:
|
||||
t.join(timeout=10)
|
||||
|
||||
def failures(self, since: float = 0.0) -> list[tuple]:
|
||||
return [e for e in self.events if e[0] >= since and e[2] not in (200, 201)]
|
||||
|
||||
def check_history(self, api: Api) -> None:
|
||||
history = api.history(self.bob, self.chat, 100)[0]
|
||||
ids, texts = [m["id"] for m in history], [m["text"] for m in history]
|
||||
assert len(ids) == len(set(ids)) and len(texts) == len(set(texts)), "duplicate after retry"
|
||||
assert {m["id"] for m in self.confirmed} <= set(ids), "a confirmed message was lost"
|
||||
|
||||
|
||||
@pytest.mark.parametrize("position", range(REPLICAS))
|
||||
def test_graceful_stop_of_each_replica_loses_no_request(api, position):
|
||||
"""SIGTERM -> readiness 503 -> HAProxy DOWN -> drain -> exit 0: not one failed request."""
|
||||
replica = sorted(running_replicas(), key=lambda r: r["name"])[position]
|
||||
with Traffic(api) as traffic:
|
||||
started = time.monotonic()
|
||||
with replica_stopped(replica, "stop"):
|
||||
stopped_after = time.monotonic() - started
|
||||
time.sleep(3)
|
||||
after = [e for e in traffic.events if e[0] > time.monotonic() - 2]
|
||||
assert after and replica["hostname"] not in {e[3] for e in after}
|
||||
time.sleep(1)
|
||||
state = json.loads(subprocess.run(["docker", "inspect", replica["id"]], capture_output=True, text=True,
|
||||
check=True).stdout)[0]["State"]
|
||||
assert traffic.failures() == [], traffic.failures()[:10]
|
||||
assert stopped_after < 15, "graceful stop must finish before stop_grace_period (no SIGKILL)"
|
||||
assert state["Status"] == "running" # restored by replica_stopped
|
||||
traffic.check_history(api)
|
||||
print(f"{replica['name']}: stop took {stopped_after:.1f} s, {len(traffic.events)} requests, 0 failed")
|
||||
|
||||
|
||||
@pytest.mark.parametrize("position", range(REPLICAS))
|
||||
def test_killed_replica_bounded_errors_recovery_and_no_duplicates(api, position):
|
||||
"""SIGKILL: requests in flight on that replica may fail (counted), new ones succeed within
|
||||
10 s, the killed replica gets nothing after detection, retries do not duplicate."""
|
||||
replica = sorted(running_replicas(), key=lambda r: r["name"])[position]
|
||||
with Traffic(api) as traffic:
|
||||
killed_at = time.monotonic()
|
||||
with replica_stopped(replica, "kill"):
|
||||
time.sleep(4)
|
||||
failures = traffic.failures(killed_at)
|
||||
last_failure = max((e[0] for e in failures), default=killed_at) - killed_at
|
||||
post_ok = [e for e in traffic.events if e[0] > killed_at and e[1] == "POST" and e[2] == 201]
|
||||
assert post_ok and post_ok[0][0] - killed_at <= 10, "writes did not recover within 10 s"
|
||||
late = [e for e in traffic.events if e[0] > killed_at + 3.5]
|
||||
assert late and replica["hostname"] not in {e[3] for e in late}
|
||||
assert last_failure <= 10
|
||||
traffic.check_history(api)
|
||||
print(f"{replica['name']}: killed; {len(failures)} failed of {len(traffic.events)}, last failure +{last_failure:.2f} s")
|
||||
|
||||
|
||||
def test_graceful_shutdown_order_in_logs(api):
|
||||
"""Logs prove the order: drain_started -> (HAProxy marks DOWN) -> shutdown_requested -> exit 0."""
|
||||
replica = sorted(running_replicas(), key=lambda r: r["name"])[-1]
|
||||
since = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime(time.time() - 1))
|
||||
with replica_stopped(replica, "stop"):
|
||||
logs = subprocess.run(["docker", "logs", "--since", since, replica["id"]], capture_output=True,
|
||||
text=True, encoding="utf-8").stdout
|
||||
lb_logs = compose("logs", "--no-log-prefix", "--since", since, "lb").stdout
|
||||
state = json.loads(subprocess.run(["docker", "inspect", replica["id"]], capture_output=True, text=True,
|
||||
check=True).stdout)[0]["State"]
|
||||
events = [json.loads(line) for line in logs.splitlines() if line.startswith("{")]
|
||||
names = [e["event"] for e in events]
|
||||
assert names.index("drain_started") < names.index("shutdown_requested") < names.index("stopped")
|
||||
t = {e["event"]: e["ts"] for e in events}
|
||||
assert t["drain_started"] < t["shutdown_requested"]
|
||||
assert state["ExitCode"] == 0 and not state["OOMKilled"]
|
||||
assert "is DOWN" in lb_logs and ("503" in lb_logs or "Connection refused" in lb_logs)
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ discovery: restart and scaling
|
||||
def test_full_restart_lb_rediscovers_all_replicas(api):
|
||||
actor = api.user("restart")
|
||||
before = {r["hostname"]: r["started_at"] for r in running_replicas()}
|
||||
run_hook("restart.sh")
|
||||
wait_healthy("postgres", "valkey", "api")
|
||||
wait_stand()
|
||||
after = {r["hostname"]: r["started_at"] for r in running_replicas()}
|
||||
assert set(after) == set(before) and all(after[h] != before[h] for h in after), "every replica restarted"
|
||||
assert set(instances_seen(api, actor["token"])) == set(after) # same token, both replicas, no LB restart needed
|
||||
|
||||
|
||||
def test_scale_up_and_down_is_picked_up_without_errors(api):
|
||||
with Traffic(api, workers=2) as traffic:
|
||||
try:
|
||||
compose("up", "-d", "--no-deps", "--no-recreate", "--wait", "--scale", f"api={REPLICAS + 1}", "api",
|
||||
timeout=180)
|
||||
lbctl.wait_up(REPLICAS + 1, timeout=30)
|
||||
actor = api.user("scale")
|
||||
assert len(set(instances_seen(api, actor["token"], 30))) == REPLICAS + 1
|
||||
finally:
|
||||
compose("up", "-d", "--no-deps", "--no-recreate", "--wait", "--scale", f"api={REPLICAS}", "api",
|
||||
timeout=180)
|
||||
wait_stand()
|
||||
time.sleep(1)
|
||||
assert traffic.failures() == [], traffic.failures()[:10]
|
||||
traffic.check_history(api)
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ level 4: deployment policy
|
||||
def test_compose_policy_limits_ports_secrets_images():
|
||||
config = json.loads(compose("config", "--format", "json").stdout)
|
||||
services = config["services"]
|
||||
assert set(services) == {"postgres", "valkey", "migrate", "api", "lb"}
|
||||
for name, svc in services.items():
|
||||
limits = svc.get("deploy", {}).get("resources", {}).get("limits", {})
|
||||
assert limits.get("cpus") and limits.get("memory"), f"{name}: CPU/RAM limits"
|
||||
assert "container_name" not in svc, f"{name}: container_name prevents scaling"
|
||||
if name in ("api", "migrate"):
|
||||
assert "build" in svc # built from this repository's source
|
||||
else:
|
||||
assert "@sha256:" in svc["image"], f"{name}: image must be pinned by digest"
|
||||
assert services["api"]["deploy"]["replicas"] >= 2 and not services["api"].get("ports")
|
||||
published = {name: svc.get("ports", []) for name, svc in services.items() if svc.get("ports")}
|
||||
assert set(published) == {"lb"} and all(p["host_ip"] in ("127.0.0.1", "::1") for p in published["lb"])
|
||||
secrets = {name: sorted(s["source"] for s in svc.get("secrets", [])) for name, svc in services.items()}
|
||||
assert secrets == {"postgres": ["postgres_password"], "valkey": ["valkey_password"],
|
||||
"migrate": ["database_url"], "api": ["cache_url", "database_url"], "lb": []}
|
||||
volumes = {name: [v["source"] for v in svc.get("volumes", []) if v["type"] == "volume"]
|
||||
for name, svc in services.items()}
|
||||
assert volumes == {"postgres": ["pgdata"], "valkey": ["valkeydata"], "migrate": [], "api": [], "lb": []}
|
||||
networks = {name: sorted(svc.get("networks", {})) for name, svc in services.items()}
|
||||
assert networks == {"postgres": ["data"], "valkey": ["data"], "migrate": ["data"], "api": ["data", "edge"],
|
||||
"lb": ["edge"]}, networks # the LB cannot reach the stores at all
|
||||
assert config["networks"]["data"].get("internal") is True
|
||||
|
||||
|
||||
def test_running_containers_have_limits_and_no_bypass_ports():
|
||||
ids = compose("ps", "-q").stdout.split()
|
||||
for item in json.loads(subprocess.run(["docker", "inspect", *ids], capture_output=True, text=True,
|
||||
check=True).stdout):
|
||||
name = item["Name"]
|
||||
assert item["HostConfig"]["NanoCpus"] > 0 and item["HostConfig"]["Memory"] > 0, name
|
||||
bindings = item["HostConfig"]["PortBindings"] or {}
|
||||
if "-lb-" in name:
|
||||
assert all(b["HostIp"] in ("127.0.0.1", "::1") for v in bindings.values() for b in v) # loopback only
|
||||
else:
|
||||
assert not bindings, f"{name} publishes ports in bypass of the LB"
|
||||
|
||||
|
||||
def test_migrations_ran_once_by_the_migrate_service():
|
||||
applied = psql("SELECT version FROM schema_migrations ORDER BY version").split()
|
||||
assert applied == ["0001_init", "0002_idempotency_keys", "0003_post_message_function"]
|
||||
migrate = compose("ps", "-a", "--format", "json", "migrate").stdout
|
||||
rows = [json.loads(line) for line in migrate.splitlines() if line.strip()]
|
||||
assert rows and all(r["State"] == "exited" and r["ExitCode"] == 0 for r in rows)
|
||||
api_logs = compose("logs", "--no-log-prefix", "api").stdout
|
||||
assert "migrate:" not in api_logs and "schema_migrations" not in api_logs # replicas never migrate
|
||||
@@ -0,0 +1,116 @@
|
||||
"""Database-level guarantees, checked over HTTP and verified directly in PostgreSQL / Valkey."""
|
||||
from __future__ import annotations
|
||||
|
||||
import threading
|
||||
import time
|
||||
import uuid
|
||||
|
||||
import pytest
|
||||
|
||||
from conftest import PASSWORD, Api, compose, psql
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
|
||||
def run_concurrently(workers: int, target) -> list:
|
||||
barrier = threading.Barrier(workers)
|
||||
results: list = [None] * workers
|
||||
|
||||
def body(index: int) -> None:
|
||||
local = Api() # one HTTP client (connection) per worker
|
||||
barrier.wait() # all workers fire at the same moment
|
||||
try:
|
||||
results[index] = target(local, index)
|
||||
except BaseException as exc: # noqa: BLE001 - reported by the test
|
||||
results[index] = exc
|
||||
|
||||
threads = [threading.Thread(target=body, args=(i,)) for i in range(workers)]
|
||||
for t in threads:
|
||||
t.start()
|
||||
for t in threads:
|
||||
t.join()
|
||||
return results
|
||||
|
||||
|
||||
def test_registration_race_exactly_one_201_and_seven_409():
|
||||
username = "race_" + uuid.uuid4().hex[:10]
|
||||
|
||||
def register(client: Api, _index: int) -> int:
|
||||
body = {"username": username, "display_name": "Гонка", "password": PASSWORD}
|
||||
response = client.client.post("/api/v1/users", json=body)
|
||||
return response.status_code
|
||||
|
||||
statuses = run_concurrently(8, register)
|
||||
# Real HTTP codes are compared: a timeout, 500 or exception fails the test instead of counting as 409.
|
||||
assert sorted(statuses, key=str) == [201] + [409] * 7, statuses
|
||||
assert psql(f"SELECT count(*) FROM users WHERE username = '{username}'") == "1"
|
||||
|
||||
|
||||
def _ins_counter() -> int:
|
||||
return int(psql("SELECT n_tup_ins FROM pg_stat_user_tables WHERE relname = 'chats'"))
|
||||
|
||||
|
||||
def test_create_chat_rolls_back_after_the_chat_row_was_inserted(api):
|
||||
"""The chat INSERT runs first; the unknown member then breaks the membership INSERT (FK).
|
||||
|
||||
Proof that the first INSERT really happened and was rolled back: PostgreSQL counts
|
||||
attempted inserts in pg_stat_user_tables.n_tup_ins even for aborted transactions,
|
||||
while the table contains no such chat and no new membership rows.
|
||||
"""
|
||||
alice = api.user("rollback")
|
||||
title = "rollback-" + uuid.uuid4().hex
|
||||
members_before = int(psql("SELECT count(*) FROM chat_members"))
|
||||
attempts_before = _ins_counter()
|
||||
error = api.call("POST", "/api/v1/chats", 400, token=alice["token"],
|
||||
json_body={"title": title, "member_ids": [alice["id"], str(uuid.uuid4())]}).json()
|
||||
assert "несуществующих" in error["error"]["message"]
|
||||
deadline = time.monotonic() + 20 # statistics are flushed by the API backend asynchronously
|
||||
while _ins_counter() == attempts_before and time.monotonic() < deadline:
|
||||
time.sleep(0.5)
|
||||
assert _ins_counter() > attempts_before, "the chat INSERT was never attempted"
|
||||
assert psql(f"SELECT count(*) FROM chats WHERE title = convert_to('{title}', 'UTF8')") == "0"
|
||||
assert int(psql("SELECT count(*) FROM chat_members")) == members_before
|
||||
|
||||
|
||||
def test_migrations_rerun_is_a_noop():
|
||||
users_before = psql("SELECT count(*) FROM users")
|
||||
result = compose("run", "--rm", "--no-deps", "migrate")
|
||||
assert "0001_init: already applied, skipped" in result.stdout
|
||||
assert "0002_idempotency_keys: already applied, skipped" in result.stdout
|
||||
assert "0003_post_message_function: already applied, skipped" in result.stdout
|
||||
assert psql("SELECT count(*) FROM schema_migrations") == "3"
|
||||
assert psql("SELECT count(*) FROM users") == users_before
|
||||
|
||||
|
||||
def test_100_messages_from_10_clients_keep_order_and_cursor_walk(api):
|
||||
clients = [api.user(f"w{i}") for i in range(10)]
|
||||
chat = api.chat(clients[0], *clients[1:])
|
||||
|
||||
def writer(client: Api, index: int) -> list[dict]:
|
||||
return [client.send(clients[index], chat, f"c{index}-m{j}") for j in range(10)]
|
||||
|
||||
results = run_concurrently(10, writer)
|
||||
assert all(isinstance(r, list) for r in results), results
|
||||
confirmed = [m for batch in results for m in batch]
|
||||
history, pages = api.history(clients[5], chat, 7)
|
||||
ids = [m["id"] for m in history]
|
||||
assert len(confirmed) == 100 and len(history) == 100 and len(set(ids)) == 100 and pages == 15
|
||||
assert set(ids) == {m["id"] for m in confirmed}
|
||||
keys = [(m["created_at"], uuid.UUID(m["id"]).int) for m in history]
|
||||
assert all(a < b for a, b in zip(keys, keys[1:])), "history is not strictly ordered"
|
||||
stamps = [m["created_at"] for m in history]
|
||||
assert len(set(stamps)) == 100, "per-chat clock must give unique timestamps"
|
||||
for i, client in enumerate(clients):
|
||||
own = [m["text"] for m in history if m["sender_id"] == client["id"]]
|
||||
assert own == [f"c{i}-m{j}" for j in range(10)]
|
||||
|
||||
|
||||
def test_secrets_are_not_stored_in_plain_text(api):
|
||||
actor = api.user("store")
|
||||
stored = psql(f"SELECT password_hash FROM users WHERE id = '{actor['id']}'")
|
||||
assert stored.startswith("$argon2id$v=19$m=19456,t=2,p=1$") and PASSWORD not in stored
|
||||
token_hex = actor["token"].encode().hex()
|
||||
keys = compose("exec", "-T", "valkey", "sh", "-c",
|
||||
'REDISCLI_AUTH="$(cat /run/secrets/valkey_password)" valkey-cli --scan --pattern "session:*"').stdout.split()
|
||||
assert keys and all(len(k) == len("session:") + 64 for k in keys)
|
||||
assert not any(actor["token"] in k or token_hex in k for k in keys)
|
||||
@@ -0,0 +1,35 @@
|
||||
"""The history index really bounds the work of a late page (>= 10 000 messages, generic plan)."""
|
||||
from __future__ import annotations
|
||||
|
||||
import re
|
||||
import sys
|
||||
|
||||
import pytest
|
||||
|
||||
from conftest import ROOT
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
sys.path.insert(0, str(ROOT / "scripts"))
|
||||
import explain_history # noqa: E402
|
||||
|
||||
|
||||
def section(text: str, number: int) -> str:
|
||||
parts = re.split(r"^=== ", text, flags=re.M)
|
||||
return next(p for p in parts if p.startswith(f"{number}."))
|
||||
|
||||
|
||||
def buffers(plan: str) -> int:
|
||||
return int(re.search(r"Buffers: shared hit=(\d+)", plan).group(1))
|
||||
|
||||
|
||||
def test_late_page_uses_index_range_not_filter():
|
||||
text = explain_history.run()
|
||||
assert "20000" in text # the target chat holds 20 000 messages
|
||||
app_late = section(text, 2)
|
||||
assert "Index Cond: ((chat_id = $1) AND (ROW(created_at, id) > ROW($2, $3)))" in app_late
|
||||
assert "Rows Removed by Filter" not in app_late and "Sort" not in app_late
|
||||
rejected = section(text, 3)
|
||||
assert "Rows Removed by Filter: 19950" in rejected
|
||||
# The app query touches a handful of pages; the rejected variants touch the whole prefix.
|
||||
assert buffers(app_late) <= 10 < buffers(rejected) and buffers(app_late) <= 10 < buffers(section(text, 4))
|
||||
@@ -0,0 +1,111 @@
|
||||
"""Dependency failures with a VALID session: bounded 503, never anonymous access, recovery afterwards.
|
||||
|
||||
Every test stops/pauses a real container and restores it in `finally` (service_down).
|
||||
Timing bounds: DB_TIMEOUT_SECONDS=3 (client deadline), CACHE_TIMEOUT_SECONDS=0.5.
|
||||
|
||||
Lab 3: requests go through HAProxy. While the outage is being detected the replicas answer
|
||||
503 themselves; once every replica fails /health/ready, HAProxy answers 503 in the same
|
||||
contract format (without X-Instance-Id). Both are "unavailable"; never 200 and never 401
|
||||
for a valid token. Liveness/readiness of the processes are checked inside each container.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import subprocess
|
||||
import time
|
||||
|
||||
import pytest
|
||||
|
||||
from conftest import ROOT, probe_inside, restart_services, running_replicas, service_down
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
DB_BOUND = 3.6 # DB_TIMEOUT_SECONDS + network/HTTP overhead
|
||||
CACHE_BOUND = 1.6 # 2 x CACHE_TIMEOUT_SECONDS + overhead
|
||||
|
||||
|
||||
def timed(api, *args, **kwargs):
|
||||
started = time.monotonic()
|
||||
response = api.call(*args, lb_error_ok=True, **kwargs)
|
||||
return response, time.monotonic() - started
|
||||
|
||||
|
||||
def processes_alive_but_not_ready() -> None:
|
||||
for replica in running_replicas():
|
||||
assert probe_inside(replica, "/health/live") == 200
|
||||
assert probe_inside(replica, "/health/ready") == 503
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def prepared(api):
|
||||
alice, bob = api.user("alice"), api.user("bob")
|
||||
chat = api.chat(alice, bob)
|
||||
message = api.send(alice, chat, "до отказа")
|
||||
return alice, bob, chat, message
|
||||
|
||||
|
||||
def check_intact(api, alice, bob, chat, message):
|
||||
"""After recovery the SAME tokens work and confirmed data is unchanged."""
|
||||
assert api.call("GET", "/api/v1/users/me", 200, token=alice["token"]).json() == alice["user"]
|
||||
items = api.call("GET", f"/api/v1/chats/{chat['id']}/messages", 200, token=bob["token"]).json()["items"]
|
||||
assert items == [message]
|
||||
|
||||
|
||||
@pytest.mark.parametrize("how", ["stop", "pause"])
|
||||
def test_valkey_failure(api, prepared, how):
|
||||
alice, bob, chat, message = prepared
|
||||
with service_down("valkey", how):
|
||||
response, took = timed(api, "GET", "/api/v1/users/me", 503, token=alice["token"])
|
||||
assert response.json()["error"]["code"] == "unavailable" and took < CACHE_BOUND, took
|
||||
api.call("GET", f"/api/v1/chats/{chat['id']}/messages", 503, token=bob["token"])
|
||||
api.call("GET", "/api/v1/users/me", (401, 503), lb_error_ok=True) # anonymous never gets access
|
||||
api.call("GET", "/health/ready", 503, lb_error_ok=True)
|
||||
processes_alive_but_not_ready()
|
||||
check_intact(api, alice, bob, chat, message)
|
||||
|
||||
|
||||
@pytest.mark.parametrize("how", ["stop", "pause"])
|
||||
def test_postgres_failure(api, prepared, how):
|
||||
alice, bob, chat, message = prepared
|
||||
with service_down("postgres", how):
|
||||
for path in ("/api/v1/users/me", f"/api/v1/chats/{chat['id']}/messages", "/api/v1/chats"):
|
||||
response, took = timed(api, "GET", path, 503, token=alice["token"])
|
||||
assert response.json()["error"]["code"] == "unavailable", path
|
||||
assert took < DB_BOUND, (path, took) # bounded even when PostgreSQL is frozen (pause)
|
||||
api.call("GET", "/api/v1/users/me", (401, 503), lb_error_ok=True) # anonymous never gets access
|
||||
api.call("GET", "/health/ready", 503, lb_error_ok=True)
|
||||
processes_alive_but_not_ready()
|
||||
check_intact(api, alice, bob, chat, message)
|
||||
|
||||
|
||||
def test_locked_table_gives_bounded_503(api, prepared):
|
||||
"""A statement stuck behind a lock is stopped by lock_timeout, not after minutes."""
|
||||
alice, bob, chat, message = prepared
|
||||
locker = subprocess.Popen(
|
||||
["docker", "compose", "exec", "-T", "postgres", "psql", "-U", "messenger", "-d", "messenger", "-c",
|
||||
"BEGIN; LOCK TABLE messages IN ACCESS EXCLUSIVE MODE; SELECT pg_sleep(6); COMMIT;"],
|
||||
cwd=ROOT, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
|
||||
try:
|
||||
time.sleep(1.5) # let the lock be taken
|
||||
response, took = timed(api, "GET", f"/api/v1/chats/{chat['id']}/messages", 503, token=bob["token"])
|
||||
assert response.json()["error"]["code"] == "unavailable" and took < DB_BOUND, took
|
||||
assert response.headers.get("x-instance-id"), "a lock is the API's own 503, the replicas stay ready"
|
||||
finally:
|
||||
locker.wait(timeout=30)
|
||||
check_intact(api, alice, bob, chat, message)
|
||||
|
||||
|
||||
def test_logout_during_valkey_outage_is_never_lost(api, prepared):
|
||||
"""Valkey is the single source of truth for sessions, so a logout cannot be 'forgotten':
|
||||
during the outage logout answers 503 (nothing revoked, the client knows it failed);
|
||||
after recovery logout works, and the revocation survives a Valkey and API restart (AOF)."""
|
||||
alice, bob, chat, message = prepared
|
||||
with service_down("valkey", "stop"):
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 503, token=alice["token"], lb_error_ok=True)
|
||||
api.call("GET", "/api/v1/users/me", 200, token=alice["token"]) # logout did not happen
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 204, token=alice["token"])
|
||||
api.call("GET", "/api/v1/users/me", 401, token=alice["token"])
|
||||
restart_services("valkey", "api")
|
||||
for _ in range(10): # repeated checks: the revoked token must never come back
|
||||
api.call("GET", "/api/v1/users/me", 401, token=alice["token"])
|
||||
api.call("POST", f"/api/v1/chats/{chat['id']}/messages", 401, token=alice["token"], json_body={"text": "x"})
|
||||
api.call("GET", "/api/v1/users/me", 200, token=bob["token"]) # other sessions survived the restart
|
||||
@@ -0,0 +1,281 @@
|
||||
"""Shared idempotency of POST /chats/{id}/messages (lab 3), through the load balancer.
|
||||
|
||||
Strict assertions: exact HTTP codes for every request, identical bodies for replays,
|
||||
the number of rows in PostgreSQL (messages AND idempotency_keys), and which replica
|
||||
answered (X-Instance-Id). Fault injection (trigger, row lock, stopped replica) is always
|
||||
undone in `finally`.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import subprocess
|
||||
import threading
|
||||
import time
|
||||
import uuid
|
||||
|
||||
import httpx2 as httpx
|
||||
import pytest
|
||||
|
||||
from conftest import BASE_URL, ROOT, Api, psql, replica_stopped, running_replicas
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
|
||||
def messages_in_db(chat: dict) -> int:
|
||||
return int(psql(f"SELECT count(*) FROM messages WHERE chat_id = '{chat['id']}'"))
|
||||
|
||||
|
||||
def key_rows(actor: dict, key: str) -> int:
|
||||
return int(psql(f"SELECT count(*) FROM idempotency_keys WHERE user_id = '{actor['id']}' AND key = '{key}'"))
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def pair(api):
|
||||
alice, bob = api.user("alice"), api.user("bob")
|
||||
return alice, bob, api.chat(alice, bob)
|
||||
|
||||
|
||||
def test_replay_returns_the_same_201_and_message_from_every_replica(api, pair):
|
||||
alice, bob, chat = pair
|
||||
key = "replay-" + uuid.uuid4().hex
|
||||
first = api.post_message(alice, chat, {"text": "один раз"}, key=key)
|
||||
served = {first.headers["x-instance-id"]}
|
||||
for _ in range(20):
|
||||
again = api.post_message(alice, chat, {"text": "один раз"}, key=key)
|
||||
assert again.status_code == 201 and again.json() == first.json()
|
||||
served.add(again.headers["x-instance-id"])
|
||||
assert len(served) >= 2, "replays must be answered by different replicas, not a local cache"
|
||||
assert api.history(bob, chat, 100)[0] == [first.json()]
|
||||
assert messages_in_db(chat) == 1 and key_rows(alice, key) == 1
|
||||
|
||||
|
||||
def test_other_body_with_the_same_key_is_409_and_keeps_the_original(api, pair):
|
||||
alice, bob, chat = pair
|
||||
key = uuid.uuid4().hex
|
||||
original = api.send(alice, chat, "оригинал", key=key)
|
||||
for body in ({"text": "другой текст"}, {"text": "оригинал "}, {"text": "Оригинал"}):
|
||||
assert api.post_message(alice, chat, body, 409, key=key).json()["error"]["code"] == "conflict"
|
||||
assert api.send(alice, chat, "оригинал", key=key) == original
|
||||
assert api.history(bob, chat, 100)[0] == [original] and key_rows(alice, key) == 1
|
||||
|
||||
|
||||
def test_key_order_and_missing_attachment_ids_are_the_same_body(api, pair):
|
||||
alice, _, chat = pair
|
||||
key = uuid.uuid4().hex
|
||||
first = api.post_message(alice, chat, None, raw='{"text":"порядок"}'.encode(), key=key).json()
|
||||
for raw in ('{"attachment_ids":[],"text":"порядок"}', '{"text":"порядок","attachment_ids":[]}',
|
||||
'{ "text" : "\\u043f\\u043e\\u0440\\u044f\\u0434\\u043e\\u043a" }'):
|
||||
assert api.post_message(alice, chat, None, raw=raw.encode(), key=key).json() == first
|
||||
assert messages_in_db(chat) == 1
|
||||
|
||||
|
||||
@pytest.mark.parametrize("value", ["", "a" * 129, "ключ", "a b", "a/b", "a,b", "a\tb", "*"])
|
||||
def test_missing_or_invalid_key_is_400(api, pair, value):
|
||||
alice, _, chat = pair
|
||||
raw_value = value.encode("utf-8")
|
||||
error = api.call("POST", f"/api/v1/chats/{chat['id']}/messages", 400, token=alice["token"],
|
||||
json_body={"text": "x"}, headers={"Idempotency-Key": raw_value} if value else {}).json()
|
||||
assert error["error"]["code"] == "invalid_request"
|
||||
assert messages_in_db(chat) == 0
|
||||
|
||||
|
||||
def test_repeated_key_header_is_400(pair):
|
||||
alice, _, chat = pair
|
||||
response = httpx.post(BASE_URL + f"/api/v1/chats/{chat['id']}/messages", json={"text": "x"}, timeout=10,
|
||||
headers=[("Authorization", "Bearer " + alice["token"]),
|
||||
("Idempotency-Key", "one"), ("Idempotency-Key", "two")])
|
||||
assert response.status_code == 400 and response.headers["x-instance-id"]
|
||||
assert messages_in_db(chat) == 0
|
||||
|
||||
|
||||
def test_longest_key_and_every_allowed_character(api, pair):
|
||||
alice, _, chat = pair
|
||||
for key in ("k" * 128, "A-Za-z0-9._:-" + uuid.uuid4().hex):
|
||||
first = api.send(alice, chat, key[:20], key=key)
|
||||
assert api.send(alice, chat, key[:20], key=key) == first
|
||||
assert messages_in_db(chat) == 2
|
||||
|
||||
|
||||
def test_scope_is_user_plus_chat_plus_key(api, pair):
|
||||
alice, bob, chat = pair
|
||||
other = api.chat(alice, bob, title="второй")
|
||||
key = "shared-key"
|
||||
a1 = api.send(alice, chat, "текст", key=key)
|
||||
b1 = api.send(bob, chat, "текст", key=key) # another user: independent
|
||||
a2 = api.send(alice, other, "текст", key=key) # another chat: independent
|
||||
assert len({a1["id"], b1["id"], a2["id"]}) == 3
|
||||
assert messages_in_db(chat) == 2 and messages_in_db(other) == 1
|
||||
|
||||
|
||||
def test_outsider_and_revoked_session_never_see_the_stored_result(api, pair):
|
||||
alice, bob, chat = pair
|
||||
eve = api.user("eve")
|
||||
key = uuid.uuid4().hex
|
||||
api.send(alice, chat, "секрет", key=key)
|
||||
for _ in range(3): # a replay checks access again: always 404, never 409 or the stored message
|
||||
assert api.post_message(eve, chat, {"text": "секрет"}, 404, key=key).json()["error"]["code"] == "not_found"
|
||||
api.post_message(eve, chat, {"text": "другое"}, 404, key=key)
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 204, token=alice["token"])
|
||||
api.post_message(alice, chat, {"text": "секрет"}, 401, key=key) # session checked again
|
||||
assert messages_in_db(chat) == 1
|
||||
|
||||
|
||||
def fire_together(workers: int, request) -> list:
|
||||
barrier = threading.Barrier(workers)
|
||||
results: list = [None] * workers
|
||||
|
||||
def body(index: int) -> None:
|
||||
client = httpx.Client(base_url=BASE_URL, timeout=15) # own connection -> LB spreads them
|
||||
barrier.wait()
|
||||
try:
|
||||
results[index] = request(client, index)
|
||||
except BaseException as exc: # noqa: BLE001 - asserted by the test
|
||||
results[index] = exc
|
||||
finally:
|
||||
client.close()
|
||||
|
||||
threads = [threading.Thread(target=body, args=(i,)) for i in range(workers)]
|
||||
for t in threads:
|
||||
t.start()
|
||||
for t in threads:
|
||||
t.join()
|
||||
return results
|
||||
|
||||
|
||||
def test_concurrent_same_key_through_different_replicas_creates_exactly_one(pair):
|
||||
alice, bob, chat = pair
|
||||
key = "race-" + uuid.uuid4().hex
|
||||
|
||||
def post(client: httpx.Client, _index: int):
|
||||
r = client.post(f"/api/v1/chats/{chat['id']}/messages", json={"text": "гонка"},
|
||||
headers={"Authorization": "Bearer " + alice["token"], "Idempotency-Key": key})
|
||||
return r.status_code, r.headers.get("x-instance-id"), r.json()
|
||||
|
||||
for _ in range(3): # three rounds: the first creates, later rounds replay concurrently
|
||||
results = fire_together(24, post)
|
||||
assert all(isinstance(r, tuple) for r in results), results
|
||||
assert [r[0] for r in results] == [201] * 24, [r[0] for r in results]
|
||||
assert len({str(r[2]) for r in results}) == 1, "every concurrent replay must return the same Message"
|
||||
assert len({r[1] for r in results}) >= 2, "requests did not reach two replicas"
|
||||
assert messages_in_db(chat) == 1 and key_rows(alice, key) == 1
|
||||
|
||||
|
||||
def test_concurrent_same_key_different_bodies_one_wins(pair):
|
||||
alice, _, chat = pair
|
||||
key = "conflict-" + uuid.uuid4().hex
|
||||
|
||||
def post(client: httpx.Client, index: int):
|
||||
r = client.post(f"/api/v1/chats/{chat['id']}/messages", json={"text": f"вариант {index % 2}"},
|
||||
headers={"Authorization": "Bearer " + alice["token"], "Idempotency-Key": key})
|
||||
return r.status_code, r.json()
|
||||
|
||||
results = fire_together(20, post)
|
||||
winners = [body for status, body in results if status == 201]
|
||||
losers = [body for status, body in results if status == 409]
|
||||
assert len(winners) + len(losers) == 20, results
|
||||
assert winners and len({str(b) for b in winners}) == 1
|
||||
assert all(b["error"]["code"] == "conflict" for b in losers)
|
||||
assert {b["text"] for b in winners} <= {"вариант 0", "вариант 1"}
|
||||
assert len(losers) == sum(1 for i in range(20) if f"вариант {i % 2}" != winners[0]["text"])
|
||||
assert messages_in_db(chat) == 1
|
||||
|
||||
|
||||
def inserts_into_keys() -> int:
|
||||
return int(psql("SELECT n_tup_ins FROM pg_stat_user_tables WHERE relname = 'idempotency_keys'"))
|
||||
|
||||
|
||||
def test_failure_after_the_key_insert_rolls_everything_back(api, pair):
|
||||
"""A trigger makes the MESSAGE insert fail AFTER the key row was inserted in the same
|
||||
transaction. Expected: 500 for the client, no key row and no message left behind
|
||||
(n_tup_ins proves the key INSERT really ran), and a retry with the same key succeeds."""
|
||||
alice, bob, chat = pair
|
||||
marker = "boom-" + uuid.uuid4().hex
|
||||
key = uuid.uuid4().hex
|
||||
psql("CREATE OR REPLACE FUNCTION lab3_test_fail() RETURNS trigger LANGUAGE plpgsql AS $$ "
|
||||
f"BEGIN IF NEW.text = convert_to('{marker}', 'UTF8') THEN RAISE EXCEPTION 'injected failure'; END IF; "
|
||||
"RETURN NEW; END $$; "
|
||||
"CREATE TRIGGER lab3_test_fail BEFORE INSERT ON messages FOR EACH ROW EXECUTE FUNCTION lab3_test_fail()")
|
||||
try:
|
||||
before = inserts_into_keys()
|
||||
error = api.post_message(alice, chat, {"text": marker}, 500, key=key)
|
||||
assert error.json()["error"]["code"] == "internal_error" and "injected" not in error.text
|
||||
assert key_rows(alice, key) == 0 and messages_in_db(chat) == 0
|
||||
deadline = time.monotonic() + 20 # table statistics are flushed asynchronously
|
||||
while inserts_into_keys() == before and time.monotonic() < deadline:
|
||||
time.sleep(0.5)
|
||||
assert inserts_into_keys() > before, "the key INSERT never ran before the failure"
|
||||
finally:
|
||||
psql("DROP TRIGGER IF EXISTS lab3_test_fail ON messages; DROP FUNCTION IF EXISTS lab3_test_fail()")
|
||||
retried = api.send(alice, chat, marker, key=key) # transient failure was not recorded
|
||||
assert api.send(alice, chat, marker, key=key) == retried
|
||||
assert api.history(bob, chat, 100)[0] == [retried] and key_rows(alice, key) == 1
|
||||
|
||||
|
||||
def lock_chat_row(chat: dict, seconds: float) -> subprocess.Popen:
|
||||
sql = (f"BEGIN; SELECT 1 FROM chats WHERE id = '{chat['id']}' FOR UPDATE; "
|
||||
f"SELECT pg_sleep({seconds}); COMMIT;")
|
||||
return subprocess.Popen(["docker", "compose", "exec", "-T", "postgres", "psql", "-U", "messenger", "-d",
|
||||
"messenger", "-c", sql], cwd=ROOT, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
|
||||
|
||||
|
||||
def wait_lock_taken(chat: dict) -> None:
|
||||
deadline = time.monotonic() + 15
|
||||
while time.monotonic() < deadline:
|
||||
if psql("SELECT count(*) FROM pg_stat_activity WHERE query LIKE '%pg_sleep%' AND state = 'active' "
|
||||
"AND pid <> pg_backend_pid()") != "0":
|
||||
return
|
||||
time.sleep(0.2)
|
||||
raise AssertionError("the row lock was not taken")
|
||||
|
||||
|
||||
def test_client_timeout_then_retry_with_the_same_key_gives_the_committed_message(api, pair):
|
||||
"""Uncertain outcome: the client gives up after 0.4 s, the server commits later (the row
|
||||
lock is held 1 s < lock_timeout). The retry must return THAT message, not a second one."""
|
||||
alice, bob, chat = pair
|
||||
key = uuid.uuid4().hex
|
||||
locker = lock_chat_row(chat, 1.0)
|
||||
try:
|
||||
wait_lock_taken(chat)
|
||||
with pytest.raises(httpx.TimeoutException):
|
||||
httpx.post(BASE_URL + f"/api/v1/chats/{chat['id']}/messages", json={"text": "неопределённо"}, timeout=0.4,
|
||||
headers={"Authorization": "Bearer " + alice["token"], "Idempotency-Key": key})
|
||||
finally:
|
||||
locker.wait(timeout=30)
|
||||
deadline = time.monotonic() + 10
|
||||
while messages_in_db(chat) == 0 and time.monotonic() < deadline:
|
||||
time.sleep(0.2)
|
||||
stored = psql(f"SELECT id FROM messages WHERE chat_id = '{chat['id']}'")
|
||||
assert stored, "the server-side transaction was expected to commit after the lock was released"
|
||||
retried = api.send(alice, chat, "неопределённо", key=key)
|
||||
assert retried["id"] == stored and messages_in_db(chat) == 1 and key_rows(alice, key) == 1
|
||||
|
||||
|
||||
def test_lock_timeout_is_a_transient_503_that_is_not_recorded(api, pair):
|
||||
"""The row lock outlives lock_timeout: the API answers 503 within its deadline and stores
|
||||
nothing; after the lock is gone the same key creates the message exactly once."""
|
||||
alice, bob, chat = pair
|
||||
key = uuid.uuid4().hex
|
||||
locker = lock_chat_row(chat, 4.0)
|
||||
try:
|
||||
wait_lock_taken(chat)
|
||||
started = time.monotonic()
|
||||
api.post_message(alice, chat, {"text": "после блокировки"}, 503, key=key)
|
||||
assert time.monotonic() - started < 3.6
|
||||
assert key_rows(alice, key) == 0 and messages_in_db(chat) == 0
|
||||
finally:
|
||||
locker.wait(timeout=30)
|
||||
created = api.send(alice, chat, "после блокировки", key=key)
|
||||
assert api.send(alice, chat, "после блокировки", key=key) == created
|
||||
assert api.history(bob, chat, 100)[0] == [created] and key_rows(alice, key) == 1
|
||||
|
||||
|
||||
def test_replay_after_the_serving_replica_is_stopped(api, pair):
|
||||
alice, bob, chat = pair
|
||||
key = uuid.uuid4().hex
|
||||
first = api.post_message(alice, chat, {"text": "до остановки"}, key=key)
|
||||
served_by = first.headers["x-instance-id"]
|
||||
replica = next(r for r in running_replicas() if r["hostname"] == served_by)
|
||||
with replica_stopped(replica):
|
||||
for _ in range(5):
|
||||
again = Api().post_message(alice, chat, {"text": "до остановки"}, key=key)
|
||||
assert again.json() == first.json() and again.headers["x-instance-id"] != served_by
|
||||
assert messages_in_db(chat) == 1
|
||||
@@ -0,0 +1,65 @@
|
||||
"""Session lifetime: expiry with a short SESSION_TTL_SECONDS, no extension by restarts,
|
||||
and the alternative SESSION_STORE=postgres. The api container is recreated with the
|
||||
override and ALWAYS recreated back with the normal configuration (api_with_env)."""
|
||||
from __future__ import annotations
|
||||
|
||||
import datetime as dt
|
||||
import time
|
||||
|
||||
import pytest
|
||||
|
||||
from conftest import api_env, api_with_env, restart_services, service_down
|
||||
|
||||
pytestmark = pytest.mark.stack
|
||||
|
||||
|
||||
def seconds_left(expires_at: str) -> float:
|
||||
moment = dt.datetime.fromisoformat(expires_at.replace("Z", "+00:00"))
|
||||
return (moment - dt.datetime.now(dt.timezone.utc)).total_seconds()
|
||||
|
||||
|
||||
def login_session(api, actor) -> dict:
|
||||
from conftest import basic
|
||||
return api.call("POST", "/api/v1/auth/sessions", 201,
|
||||
headers={"Authorization": basic(actor["username"], actor["password"])}, schema="Session").json()
|
||||
|
||||
|
||||
@pytest.mark.parametrize("store", ["valkey", "postgres"])
|
||||
def test_short_ttl_expires(api, store):
|
||||
with api_with_env(SESSION_TTL_SECONDS="2", SESSION_STORE=store):
|
||||
assert api_env("SESSION_TTL_SECONDS") == "2" # the contract variable reaches the container
|
||||
actor = api.register("ttl")
|
||||
session = login_session(api, actor)
|
||||
assert 1.0 < seconds_left(session["expires_at"]) <= 2.1
|
||||
api.call("GET", "/api/v1/users/me", 200, token=session["token"])
|
||||
time.sleep(max(0.0, seconds_left(session["expires_at"])) + 0.3)
|
||||
api.call("GET", "/api/v1/users/me", 401, token=session["token"])
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 401, token=session["token"])
|
||||
|
||||
|
||||
def test_restart_does_not_extend_ttl(api):
|
||||
with api_with_env(SESSION_TTL_SECONDS="12"):
|
||||
actor = api.register("ttlrestart")
|
||||
session = login_session(api, actor)
|
||||
restart_services("valkey", "api")
|
||||
assert seconds_left(session["expires_at"]) > 0, "restart took longer than the TTL; increase it"
|
||||
api.call("GET", "/api/v1/users/me", 200, token=session["token"])
|
||||
time.sleep(max(0.0, seconds_left(session["expires_at"])) + 0.3)
|
||||
api.call("GET", "/api/v1/users/me", 401, token=session["token"])
|
||||
|
||||
|
||||
def test_postgres_session_store(api):
|
||||
"""SESSION_STORE=postgres: sessions survive an API restart, logout revokes, and a DB outage
|
||||
is 503 for a valid token (not 401 and not anonymous access)."""
|
||||
with api_with_env(SESSION_STORE="postgres"):
|
||||
assert api_env("SESSION_STORE") == "postgres"
|
||||
actor = api.user("pgstore")
|
||||
restart_services("api")
|
||||
api.call("GET", "/api/v1/users/me", 200, token=actor["token"])
|
||||
with service_down("postgres", "stop"):
|
||||
assert api.call("GET", "/api/v1/users/me", 503, token=actor["token"],
|
||||
lb_error_ok=True).json()["error"]["code"] == "unavailable"
|
||||
api.call("GET", "/api/v1/users/me", (401, 503), lb_error_ok=True)
|
||||
api.call("GET", "/api/v1/users/me", 200, token=actor["token"])
|
||||
api.call("DELETE", "/api/v1/auth/sessions/current", 204, token=actor["token"])
|
||||
api.call("GET", "/api/v1/users/me", 401, token=actor["token"])
|
||||
@@ -0,0 +1,202 @@
|
||||
"""Pure functions, no services: validation, cursor, Authorization parsing, Argon2id, config."""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import base64
|
||||
import datetime as dt
|
||||
|
||||
import pytest
|
||||
|
||||
from messenger import authhdr, cursor, passwords, validation
|
||||
from messenger.config import ConfigError, Settings
|
||||
from messenger.errors import InvalidRequest, Unauthorized
|
||||
from messenger.timefmt import UTC, from_us, rfc3339, to_us
|
||||
|
||||
|
||||
def b64(raw: bytes) -> str:
|
||||
return base64.b64encode(raw).decode()
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ password rules
|
||||
@pytest.mark.parametrize("password", ["Twelve-chars", "я" * 128, "пароль:с:двоеточиями", "😀" * 12, "a b c d e f g"])
|
||||
def test_valid_passwords(password):
|
||||
assert validation.password_field({"password": password}) == password
|
||||
|
||||
|
||||
@pytest.mark.parametrize("password", ["short-11ch", "я" * 129, " " * 12, "\t\n " * 5, "\ud800" * 12, None, 123456789012])
|
||||
def test_invalid_passwords(password):
|
||||
with pytest.raises(InvalidRequest):
|
||||
validation.password_field({"password": password})
|
||||
|
||||
|
||||
def test_text_may_contain_nul_but_not_lone_surrogates():
|
||||
assert validation.text_field({"t": "a\u0000b"}, "t", 10) == "a\u0000b"
|
||||
with pytest.raises(InvalidRequest):
|
||||
validation.text_field({"t": "a\ud800"}, "t", 10)
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ cursor
|
||||
CHAT = "3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11"
|
||||
|
||||
|
||||
def test_cursor_roundtrip_and_binding():
|
||||
key = (to_us(dt.datetime(2026, 10, 12, 12, 0, 0, 123456, tzinfo=UTC)), "11111111-1111-4111-8111-111111111111")
|
||||
token = cursor.encode(cursor.MESSAGES, CHAT, key)
|
||||
assert cursor.decode(token, cursor.MESSAGES, CHAT) == key
|
||||
with pytest.raises(InvalidRequest):
|
||||
cursor.decode(token, cursor.MESSAGES, "00000000-0000-4000-8000-000000000000") # other chat
|
||||
with pytest.raises(InvalidRequest):
|
||||
cursor.decode(token, cursor.CHATS, CHAT) # other resource
|
||||
|
||||
|
||||
@pytest.mark.parametrize("payload", [
|
||||
b"v1|m|" + CHAT.encode() + b"|253402300800000000|11111111-1111-4111-8111-111111111111", # 1 us after 9999-12-31
|
||||
b"v1|m|" + CHAT.encode() + b"|9999999999999999999|11111111-1111-4111-8111-111111111111", # int64 edge
|
||||
b"v1|m|" + CHAT.encode() + b"|-1|11111111-1111-4111-8111-111111111111",
|
||||
b"v1|m|" + CHAT.encode() + b"|1|not-a-uuid",
|
||||
b"v1|m|" + CHAT.encode() + b"|1|11111111-1111-4111-8111-11111111111G",
|
||||
b"v2|m|" + CHAT.encode() + b"|1|11111111-1111-4111-8111-111111111111",
|
||||
])
|
||||
def test_broken_cursors_are_400_not_500(payload):
|
||||
token = base64.urlsafe_b64encode(payload).rstrip(b"=").decode()
|
||||
with pytest.raises(InvalidRequest):
|
||||
cursor.decode(token, cursor.MESSAGES, CHAT)
|
||||
|
||||
|
||||
def test_time_helpers_are_exact():
|
||||
moment = dt.datetime(2026, 2, 28, 23, 59, 59, 999999, tzinfo=UTC)
|
||||
assert from_us(to_us(moment)) == moment
|
||||
assert rfc3339(moment) == "2026-02-28T23:59:59.999999Z"
|
||||
assert from_us(cursor.MAX_TS_US).year == 9999
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ Authorization header
|
||||
def test_basic_is_parsed_and_password_may_contain_colons():
|
||||
assert authhdr.parse_basic("Basic " + b64("alice:pa:ss:wörd".encode())) == ("alice", "pa:ss:wörd")
|
||||
assert authhdr.parse_basic("basic " + b64(b"bob_1:x" * 1)) == ("bob_1", "x")
|
||||
|
||||
|
||||
@pytest.mark.parametrize("header", [
|
||||
None, "", "Basic", "Basic é", "Basic !!!!", "Basic " + b64(b"no-colon"), "Basic " + b64(b"\xff\xfe:pw"),
|
||||
"Basic " + b64(b":password"), "Basic " + b64(b"alice:"), "Basic " + b64("Алиса:pw".encode()),
|
||||
"Bearer " + "a" * 43, "Basic " + "A" * 2000,
|
||||
])
|
||||
def test_bad_basic_is_401_with_basic_challenge(header):
|
||||
with pytest.raises(Unauthorized) as info:
|
||||
authhdr.parse_basic(header)
|
||||
assert info.value.headers["WWW-Authenticate"].startswith("Basic realm=\"messenger\"")
|
||||
|
||||
|
||||
@pytest.mark.parametrize("header", [None, "", "Bearer", "Bearer short", "Bearer " + "a" * 43 + "!",
|
||||
"Basic " + b64(b"alice:password-123"), "Token " + "a" * 43, "Bearer " + "é" * 43])
|
||||
def test_bad_bearer_is_401_with_bearer_challenge(header):
|
||||
with pytest.raises(Unauthorized) as info:
|
||||
authhdr.parse_bearer(header)
|
||||
assert info.value.headers["WWW-Authenticate"] == 'Bearer realm="messenger"'
|
||||
|
||||
|
||||
def test_bearer_ok():
|
||||
token = "A_b-" + "x" * 39
|
||||
assert authhdr.parse_bearer("Bearer " + token) == token
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ Argon2id
|
||||
def test_argon2id_hash_and_verify():
|
||||
async def scenario():
|
||||
stored = await passwords.hash_password("я" * 128)
|
||||
assert stored.startswith("$argon2id$v=19$m=19456,t=2,p=1$")
|
||||
assert await passwords.verify_password(stored, "я" * 128)
|
||||
assert not await passwords.verify_password(stored, "я" * 127)
|
||||
assert not await passwords.verify_password(None, "anything") # unknown user, dummy hash
|
||||
assert not await passwords.verify_password("not-a-hash", "anything")
|
||||
assert await passwords.hash_password("same password!") != await passwords.hash_password("same password!") # salt
|
||||
asyncio.run(scenario())
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ config
|
||||
def test_settings_read_secret_files_and_contract_names(tmp_path, monkeypatch):
|
||||
url_file = tmp_path / "database_url"
|
||||
url_file.write_text("postgresql://u:p@db/x\n", encoding="utf-8")
|
||||
monkeypatch.setenv("DATABASE_URL_FILE", str(url_file))
|
||||
monkeypatch.setenv("SESSION_STORE", "redis")
|
||||
monkeypatch.setenv("CACHE_URL", "redis://:p@valkey:6379/0")
|
||||
monkeypatch.setenv("SESSION_TTL_SECONDS", "2")
|
||||
settings = Settings.from_env()
|
||||
assert settings.database_url == "postgresql://u:p@db/x"
|
||||
assert settings.session_store == "valkey" and settings.session_ttl_seconds == 2
|
||||
monkeypatch.setenv("SESSION_TTL_SECONDS", "0")
|
||||
with pytest.raises(ConfigError):
|
||||
Settings.from_env()
|
||||
monkeypatch.setenv("SESSION_TTL_SECONDS", "2")
|
||||
monkeypatch.setenv("DATABASE_URL_FILE", str(tmp_path / "missing"))
|
||||
with pytest.raises(ConfigError) as info:
|
||||
Settings.from_env()
|
||||
assert "u:p" not in str(info.value)
|
||||
|
||||
|
||||
# ------------------------------------------------------------------ lab 3: idempotency, JSON, instance id
|
||||
@pytest.mark.parametrize("key", ["a", "k" * 128, "A-Za-z0-9._:-", "2026-10-16T10:00:00.1"])
|
||||
def test_valid_idempotency_keys(key):
|
||||
assert validation.idempotency_key(key) == key
|
||||
|
||||
|
||||
@pytest.mark.parametrize("key", [None, "", "k" * 129, "ключ", "a b", "a/b", "a,b", "a\tb", "a\x00b", "*", "é"])
|
||||
def test_invalid_idempotency_keys(key):
|
||||
with pytest.raises(InvalidRequest):
|
||||
validation.idempotency_key(key)
|
||||
|
||||
|
||||
@pytest.mark.parametrize("raw", [b'{"text":"a","text":"b"}', b'{"a":{"x":1,"x":2}}', b'\xff', b'{"text":"x"} {}'])
|
||||
def test_duplicate_keys_and_garbage_json_are_rejected(raw):
|
||||
with pytest.raises(InvalidRequest):
|
||||
validation.parse_json_object(raw)
|
||||
|
||||
|
||||
def test_attachment_ids_absent_equals_empty_and_is_validated():
|
||||
assert validation.attachment_ids_field({"text": "x"}) == []
|
||||
assert validation.attachment_ids_field({"attachment_ids": []}) == []
|
||||
one = "11111111-1111-4111-8111-111111111111"
|
||||
assert validation.attachment_ids_field({"attachment_ids": [one.upper()]}) == [one]
|
||||
for bad in (None, "x", [1], ["nope"], [one, one.upper()], [one] * 11):
|
||||
with pytest.raises(InvalidRequest):
|
||||
validation.attachment_ids_field({"attachment_ids": bad})
|
||||
|
||||
|
||||
def test_request_fingerprint_ignores_key_order_and_default_attachments():
|
||||
from messenger.app import request_fingerprint
|
||||
assert request_fingerprint("привет", []) == request_fingerprint("привет", [])
|
||||
assert request_fingerprint("привет", []) != request_fingerprint("привет ", [])
|
||||
assert request_fingerprint("a\u0000b", []) != request_fingerprint("ab", [])
|
||||
one = "11111111-1111-4111-8111-111111111111"
|
||||
assert request_fingerprint("x", [one]) != request_fingerprint("x", [])
|
||||
|
||||
|
||||
def test_instance_id_defaults_to_hostname_and_is_validated(monkeypatch):
|
||||
import socket
|
||||
monkeypatch.setenv("DATABASE_URL", "postgresql://x@y/z")
|
||||
monkeypatch.delenv("INSTANCE_ID", raising=False)
|
||||
assert Settings.from_env().instance_id == socket.gethostname()
|
||||
monkeypatch.setenv("INSTANCE_ID", "replica-a")
|
||||
assert Settings.from_env().instance_id == "replica-a"
|
||||
for bad in ("with space", "x" * 65, "ip/1"):
|
||||
monkeypatch.setenv("INSTANCE_ID", bad)
|
||||
with pytest.raises(ConfigError):
|
||||
Settings.from_env()
|
||||
|
||||
|
||||
def test_idempotency_ttl_cannot_be_below_24_hours(monkeypatch):
|
||||
monkeypatch.setenv("DATABASE_URL", "postgresql://x@y/z")
|
||||
monkeypatch.setenv("IDEMPOTENCY_TTL_SECONDS", "3600")
|
||||
with pytest.raises(ConfigError):
|
||||
Settings.from_env()
|
||||
|
||||
|
||||
def test_message_write_path_setting(monkeypatch):
|
||||
monkeypatch.setenv("DATABASE_URL", "postgresql://x@y/z")
|
||||
monkeypatch.delenv("MESSAGE_WRITE_PATH", raising=False)
|
||||
assert Settings.from_env().message_write_path == "function"
|
||||
monkeypatch.setenv("MESSAGE_WRITE_PATH", "multi")
|
||||
assert Settings.from_env().message_write_path == "multi"
|
||||
monkeypatch.setenv("MESSAGE_WRITE_PATH", "fast")
|
||||
with pytest.raises(ConfigError):
|
||||
Settings.from_env()
|
||||
Reference in New Issue
Block a user