新增:Linux Docker 部署与自动回收、备份及容量监测

This commit is contained in:
2026-11-10 14:53:07 +08:00
parent 3962521a5c
commit dc5b9b76ec
45 changed files with 5591 additions and 35 deletions
+88
View File
@@ -0,0 +1,88 @@
#!/usr/bin/env python3
"""与后端共享 PID 命名空间的低优先级维护任务;不向容器暴露 Docker socket。"""
import json
import logging
import os
from pathlib import Path
import signal
import sys
import time
import shutil
ROOT = Path(__file__).resolve().parents[2]
sys.path[:0] = [str(ROOT / 'scripts'), str(ROOT / 'deploy/macos')]
import backup_retention as br
import operations
import retention
def console_notify(title, message):
print(json.dumps({'event': 'capacity-alert', 'title': title, 'message': message}, ensure_ascii=False), flush=True)
return {'submitted': True, 'channel': 'container-log'}
def snapshot(root, support=None, now=None):
import backup
import managed_test_resources as resources
import log_retention
root = Path(root); now = time.time() if now is None else now
cfg = json.loads((root / '.runtime/local-config.json').read_text())
rows = backup.client('SELECT TABLE_SCHEMA,COALESCE(SUM(DATA_LENGTH+INDEX_LENGTH),0) FROM information_schema.tables GROUP BY TABLE_SCHEMA;')
estimates = {parts[0]: int(parts[1]) for line in rows.splitlines() if len(parts := line.split('\t')) == 2}
registered = 0; overdue = 0
for path in resources.registry(root).glob('*.json'):
item = resources.load(root, path.stem)
if not item or not item.get('created'): continue
registered += estimates.get(item['database'], 0)
if now >= item.get('expiresAt', now + 1): overdue += 1
private = sum(operations.allocated(p) for p in (root / '.runtime/retention').glob('verify-*') if p.is_dir() and not p.is_symlink())
mysql = estimates.get(cfg['mysqlDatabase'], 0)
archive = operations.allocated(root / 'runtimedata/archive'); backups = operations.allocated(root / 'output/backups')
native = operations.allocated(root / '.runtime/nao-users')
logs = sum(p.stat().st_blocks * 512 for p in log_retention.candidates(root, root / '.runtime/operations'))
core = mysql + archive + backups + native + logs
free = min(shutil.disk_usage(root / p).free for p in ('.runtime', 'output/backups', 'output/logs', 'runtimedata/archive'))
return {'at': now, 'mysqlBytes': mysql, 'mysqlMeasurement': 'information_schema_estimate_excludes_system_redo_undo',
'archiveBytes': archive, 'backupBytes': backups, 'logBytes': logs, 'nativeDataBytes': native,
'registeredTestBytes': registered, 'privateVerifierBytes': private, 'temporaryBytes': registered + private,
'coreBytes': core, 'artifactBytes': 0, 'permanentBytes': core, 'freeBytes': free,
'overdueTestResources': overdue}
def once(role):
if role == 'monitor':
operations.snapshot = snapshot
return operations.run(ROOT, support=ROOT / '.runtime/operations', notifier=console_notify)
return retention.run(ROOT)
def main():
role = sys.argv[1]
if role not in ('retention', 'monitor'): raise ValueError('Unknown worker role')
try: os.nice(10)
except OSError: pass
logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')
stopping = False
def stop(signum, frame):
nonlocal stopping
stopping = True
signal.signal(signal.SIGTERM, stop); signal.signal(signal.SIGINT, stop)
while not stopping:
try:
once(role)
except (br.Busy, retention.Busy): pass
except Exception as error:
if role == 'monitor': operations.failed_run(ROOT, error, notifier=console_notify)
else:
path = ROOT / '.runtime/retention/state.json'
state = json.loads(path.read_text()) if path.exists() else {}
state['consecutiveFailures'] = min(1000, state.get('consecutiveFailures', 0) + 1)
state['lastError'] = {'time': time.time(), 'type': type(error).__name__}
br.atomic(path, state)
print(json.dumps({'event': 'maintenance-error', 'role': role, 'errorType': type(error).__name__}), flush=True)
for _ in range(300):
if stopping: break
time.sleep(1)
if __name__ == '__main__': main()