Files
mtusi2026lab2/scripts/track1_summary.py
T

105 lines
5.2 KiB
Python

#!/usr/bin/env python3
"""Сводка evidence/track1/*.json в Markdown: медиана по повторам и таймлайн отказов."""
import json
import re
import statistics
import sys
from pathlib import Path
sys.stdout.reconfigure(encoding='utf-8')
out = Path(sys.argv[1] if len(sys.argv) > 1 else 'evidence/track1')
runs = {}
for f in sorted(out.glob('*.json')):
data = json.loads(f.read_text(encoding='utf-8'))
group = re.sub(r'-r\d+$', '', data['label'])
runs.setdefault(group, []).append(data)
def med(rows, key, sub=None):
vals = [(r.get(sub) or {}).get(key, 0) if sub else r.get(key, 0) for r in rows]
return statistics.median(vals) if vals else 0
def hit_ratio(rows):
vals = []
for r in rows:
c = r.get('cache') or {}
total = c.get('keyspace_hits', 0) + c.get('keyspace_misses', 0)
if total:
vals.append(c['keyspace_hits'] / total)
return '%.1f%%' % (100 * statistics.median(vals)) if vals else '—'
def resources(group):
"""Средний CPU% и максимум RAM (MiB) по сервисам из *.stats.txt прогонов группы."""
cpu, mem = {}, {}
for f in out.glob(group + '*.stats.txt'):
if re.sub(r'-r\d+$', '', f.name[:-len('.stats.txt')]) != group:
continue
for line in f.read_text(encoding='utf-8', errors='replace').splitlines():
parts = line.split(';')
m = re.match(r'mtusi-lab2-(api|postgres|valkey|valkey-cache)-1$', parts[0]) if len(parts) == 3 else None
if not m:
continue
svc = m.group(1)
cpu.setdefault(svc, []).append(float(parts[1].rstrip('%') or 0))
used = parts[2].split('/')[0].strip()
num = float(re.sub(r'[^0-9.]', '', used) or 0)
mib = num / 1024 if used.endswith('KiB') else num * 1024 if used.endswith('GiB') else num
mem[svc] = max(mem.get(svc, 0), mib)
return {s: (statistics.mean(v), mem.get(s, 0)) for s, v in cpu.items()}
def res_cell(group, svc):
r = resources(group).get(svc)
return '%.0f%% / %.0f MiB' % r if r else '—'
def ratio(rows, bad, total):
return '%d / %d' % (sum(r.get(bad, 0) for r in rows), sum(r.get(total, 0) for r in rows))
order = ['off-d0', 'naive-d0', 'safe-d0', 'naive-d20', 'safe-d20']
print('## Трек 1: медианы по повторам\n')
print('| Режим | Повт. | Чтений/с | p50, мс | p95, мс | p99, мс | SQL страницы | SQL членство | SQL на чтение | Hit ratio | RAW писателя: нарушений / проверок | Stale reads: нарушений / проверок | Ошибки |')
print('|---|---|---|---|---|---|---|---|---|---|---|---|---|')
for g in order:
rows = runs.get(g)
if not rows:
continue
errors = sum(r['read_errors'] + r['write_errors'] for r in rows)
print('| %s | %d | %.0f | %.2f | %.2f | %.2f | %.0f | %.0f | %.4f | %s | %s | %s | %d |' % (
g, len(rows), med(rows, 'read_rps'), med(rows, 'read_p50_ms'), med(rows, 'read_p95_ms'),
med(rows, 'read_p99_ms'), med(rows, 'history_page_calls', 'sql'), med(rows, 'membership_calls', 'sql'),
med(rows, 'sql_read_path_per_read'), hit_ratio(rows),
ratio(rows, 'raw_violations', 'raw_checks'), ratio(rows, 'stale_reads', 'stale_checks'), errors))
print('\nCPU (среднее, 100% = 1 ядро) / RAM (максимум) по контейнерам за прогоны:\n')
print('| Режим | api | postgres | valkey (сессии) | valkey-cache |')
print('|---|---|---|---|---|')
for g in order:
if g in runs:
print('| %s | %s | %s | %s | %s |' % (g, res_cell(g, 'api'), res_cell(g, 'postgres'),
res_cell(g, 'valkey'), res_cell(g, 'valkey-cache')))
print('\nКаждый прогон по отдельности:\n')
for g in order:
for r in runs.get(g, []):
print('- %s: чтений %d, p95 %.1f мс, SQL страницы %d, RAW %d/%d, stale reads %d/%d' % (
r['label'], r['reads_ok'], r['read_p95_ms'], (r.get('sql') or {}).get('history_page_calls', 0),
r['raw_violations'], r['raw_checks'], r.get('stale_reads', 0), r.get('stale_checks', 0)))
for g in ('safe-restart', 'safe-pause'):
for r in runs.get(g, []):
print('\n## %s: таймлайн (кэш недоступен с 10-й по ~15-ю секунду)\n' % g)
print('Итого: чтений %d, ошибок чтения %d, ошибок записи %d, RAW %d/%d, stale reads %d/%d, SQL страницы %d.\n' % (
r['reads_ok'], r['read_errors'], r['write_errors'], r['raw_violations'], r['raw_checks'],
r.get('stale_reads', 0), r.get('stale_checks', 0), (r.get('sql') or {}).get('history_page_calls', 0)))
print('| с | чтений | ошибок чт. | p95, мс | записей | ошибок зап. | RAW нарушений | stale reads |')
print('|---|---|---|---|---|---|---|---|')
for s in r['timeline']:
print('| %d | %d | %d | %.1f | %d | %d | %d | %d |' % (
s['s'], s['reads'], s['read_errors'], s['read_p95_ms'], s['writes'], s['write_errors'],
s['raw_violations'], s.get('stale_reads', 0)))