Поставить реплики API за HAProxy и описать остановку одной из них.

Внешний порт остаётся только у балансировщика, миграции применяются один раз, а stop-one оставляет выбранную реплику остановленной.
This commit is contained in:
2026-11-07 12:46:37 +03:00
parent 3601ca5443
commit fd54119d3f
8 changed files with 301 additions and 1 deletions
+15 -1
View File
@@ -3,11 +3,25 @@
APP_HOST=0.0.0.0
APP_PORT=8080
APP_ENV=lab
LOG_LEVEL=info
SHUTDOWN_GRACE_SECONDS=10
SESSION_TTL_SECONDS=3600
CACHE_TTL_SECONDS=60
# Choose credentials locally. File paths contain no secret values.
DATABASE_URL_FILE=/run/secrets/database_url
# Optional on 3, required use on 4 in lab2; Redis OR Valkey:
SESSION_STORE=postgres
SESSION_STORE=redis
CACHE_URL_FILE=/run/secrets/cache_url
API_REPLICAS=2
API_CPUS=0.50
API_MEM=256m
POSTGRES_CPUS=1.0
POSTGRES_MEM=512m
DB_POOL_MAX=4
LB_DRAIN_SECONDS=2
IDEMPOTENCY_TTL_SECONDS=86400
# Compose interpolates these. Synthetic local values only; replace after copy.
POSTGRES_PASSWORD=lab2-dev-only-change-me
DATABASE_URL=postgresql://messenger:lab2-dev-only-change-me@postgres:5432/messenger
REDIS_URL=redis://redis:6379/0
+131
View File
@@ -0,0 +1,131 @@
# Client -> lb:8080 -> api (scale) -> shared Postgres and Redis.
# API replicas do not publish host ports. Migrations run once in `migrate`.
# Check period: HAProxy inter 1s, fall 2, rise 1. Worst detection of a dead
# or not-ready replica is about 2s, plus the 1s DNS hold. API on SIGTERM
# fails /health/ready immediately and exits after LB_DRAIN_SECONDS (default 2).
services:
migrate:
build:
context: .
dockerfile: Dockerfile
image: mtusi-messenger:lab3
command: ["python", "-m", "app.migrate"]
environment:
APP_ENV: lab
LOG_LEVEL: info
DATABASE_URL: postgresql://messenger:${POSTGRES_PASSWORD}@postgres:5432/messenger
DB_POOL_MIN: "1"
DB_POOL_MAX: "2"
MIGRATIONS_ON_START: "1"
depends_on:
postgres:
condition: service_healthy
restart: "no"
mem_limit: 256m
cpus: "0.50"
api:
build:
context: .
dockerfile: Dockerfile
image: mtusi-messenger:lab3
environment:
APP_HOST: 0.0.0.0
APP_PORT: "8080"
APP_ENV: lab
LOG_LEVEL: info
SHUTDOWN_GRACE_SECONDS: "3"
LB_DRAIN_SECONDS: ${LB_DRAIN_SECONDS:-2}
SESSION_TTL_SECONDS: ${SESSION_TTL_SECONDS:-3600}
CACHE_TTL_SECONDS: ${CACHE_TTL_SECONDS:-60}
IDEMPOTENCY_TTL_SECONDS: ${IDEMPOTENCY_TTL_SECONDS:-86400}
SESSION_STORE: redis
DATABASE_URL: postgresql://messenger:${POSTGRES_PASSWORD}@postgres:5432/messenger
REDIS_URL: redis://redis:6379/0
DB_POOL_MIN: "1"
DB_POOL_MAX: ${DB_POOL_MAX:-4}
DB_COMMAND_TIMEOUT: "5"
DB_CONNECT_TIMEOUT: "5"
MIGRATIONS_ON_START: "0"
depends_on:
migrate:
condition: service_completed_successfully
postgres:
condition: service_healthy
redis:
condition: service_healthy
expose:
- "8080"
restart: unless-stopped
mem_limit: ${API_MEM:-256m}
cpus: ${API_CPUS:-0.50}
stop_grace_period: 8s
healthcheck:
test:
[
"CMD",
"python",
"-c",
"import urllib.request; urllib.request.urlopen('http://127.0.0.1:8080/health/ready', timeout=2)",
]
interval: 5s
timeout: 3s
retries: 6
start_period: 5s
lb:
image: haproxy:2.8.14-alpine
volumes:
- ./infra/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro
ports:
- "127.0.0.1:8080:8080"
depends_on:
api:
condition: service_started
restart: unless-stopped
mem_limit: 128m
cpus: "0.25"
postgres:
image: postgres:16.6-alpine
environment:
POSTGRES_USER: messenger
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
POSTGRES_DB: messenger
volumes:
- pg_data:/var/lib/postgresql/data
healthcheck:
test: ["CMD-SHELL", "pg_isready -U messenger -d messenger"]
interval: 3s
timeout: 3s
retries: 20
restart: unless-stopped
mem_limit: ${POSTGRES_MEM:-512m}
cpus: ${POSTGRES_CPUS:-1.0}
redis:
image: redis:7.4.2-alpine
command:
- redis-server
- --appendonly
- "yes"
- --appendfsync
- everysec
- --save
- "60"
- "1"
volumes:
- redis_data:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 3s
timeout: 3s
retries: 20
restart: unless-stopped
mem_limit: 128m
cpus: "0.50"
volumes:
pg_data:
redis_data:
+35
View File
@@ -0,0 +1,35 @@
global
log stdout format raw local0 info
maxconn 200
defaults
log global
mode http
option httplog
timeout connect 2s
timeout client 10s
timeout server 10s
timeout http-request 5s
timeout check 2s
timeout queue 2s
retries 1
resolvers docker
nameserver dns 127.0.0.11:53
resolve_retries 3
timeout resolve 1s
timeout retry 1s
hold valid 1s
accepted_payload_size 8192
frontend fe_http
bind *:8080
default_backend be_api
backend be_api
balance roundrobin
option httpchk
http-check send meth GET uri /health/ready ver HTTP/1.1 hdr host api
http-check expect status 200
default-server inter 1s fall 2 rise 1
server-template srv 8 api:8080 check resolvers docker init-addr none resolve-prefer ipv4
+3
View File
@@ -0,0 +1,3 @@
@echo off
python "%~dp0restart.py"
exit /b %ERRORLEVEL%
+20
View File
@@ -0,0 +1,20 @@
#!/usr/bin/env python3
"""Restart API, PostgreSQL and Redis without removing named volumes."""
from pathlib import Path
import subprocess
import sys
ROOT = Path(__file__).resolve().parents[1]
def main() -> int:
env_file = ROOT / ".env"
cmd = ["docker", "compose"]
if env_file.is_file():
cmd.extend(["--env-file", str(env_file)])
cmd.extend(["restart", "postgres", "redis", "api"])
return subprocess.run(cmd, cwd=ROOT).returncode
if __name__ == "__main__":
sys.exit(main())
+10
View File
@@ -0,0 +1,10 @@
#!/bin/sh
# Restart API replicas and every store they use. Volumes are kept.
set -eu
ROOT=$(CDPATH= cd -- "$(dirname -- "$0")/.." && pwd)
cd "$ROOT"
if [ -f .env ]; then
docker compose --env-file .env restart postgres redis api
else
docker compose restart postgres redis api
fi
+55
View File
@@ -0,0 +1,55 @@
#!/bin/sh
# Replace API replicas one at a time. The other replica keeps serving.
set -eu
ROOT=$(CDPATH= cd -- "$(dirname -- "$0")/.." && pwd)
cd "$ROOT"
compose() {
if [ -f .env ]; then
docker compose --env-file .env "$@"
else
docker compose "$@"
fi
}
if command -v python3 >/dev/null 2>&1; then
PY=python3
else
PY=python
fi
ids=$(compose ps -q api)
count=0
for id in $ids; do
count=$((count + 1))
done
if [ "$count" -lt 2 ]; then
echo "rolling update needs at least 2 running API replicas" >&2
exit 1
fi
for id in $ids; do
echo "replacing $(docker inspect -f '{{.Config.Hostname}}' "$id")"
docker stop -t 8 "$id" >/dev/null
docker rm "$id" >/dev/null
compose up -d --scale "api=$count" --no-deps --no-recreate api >/dev/null
ready=0
i=0
while [ "$i" -lt 40 ]; do
running=0
for cid in $(compose ps -q api); do
running=$((running + 1))
done
if [ "$running" -ge "$count" ] && "$PY" -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8080/health/ready', timeout=2)" >/dev/null 2>&1; then
ready=1
break
fi
i=$((i + 1))
sleep 1
done
if [ "$ready" -ne 1 ]; then
echo "replacement did not become ready" >&2
exit 1
fi
done
echo "rolling update finished; replicas=$count"
+32
View File
@@ -0,0 +1,32 @@
#!/bin/sh
# Stop the single API replica whose X-Instance-Id is TARGET_INSTANCE_ID.
# Leaves it stopped. Does not touch the database, Redis, the load balancer, or volumes.
set -eu
: "${TARGET_INSTANCE_ID:?test provides target API instance alias}"
ROOT=$(CDPATH= cd -- "$(dirname -- "$0")/.." && pwd)
cd "$ROOT"
compose() {
if [ -f .env ]; then
docker compose --env-file .env "$@"
else
docker compose "$@"
fi
}
found=0
for id in $(compose ps -q api); do
iid=$(docker exec "$id" cat /tmp/instance-id 2>/dev/null | tr -d '\r\n' || true)
host=$(docker inspect -f '{{.Config.Hostname}}' "$id")
if [ "$iid" = "$TARGET_INSTANCE_ID" ] || [ "$host" = "$TARGET_INSTANCE_ID" ]; then
docker stop -t 5 "$id" >/dev/null
found=1
break
fi
done
if [ "$found" -ne 1 ]; then
echo "no running API replica matched TARGET_INSTANCE_ID" >&2
exit 1
fi