Files
awoooi/apps
OG T 5b9b36f30d
All checks were successful
CD Pipeline / build-and-deploy (push) Successful in 14m31s
fix(ci)+feat(aiops): cd.yaml shared network + rule_catalog_sync (ADR-090 E3)
CI 修復 (c0f3509 第三次 fail 真因):
  c0f3509 log: 'Detected act task network: (none, will fall back to bridge)'
  → grep ACT_NET 在 CI 環境未 match → fallback bridge
  → default bridge 不支援 container name DNS → pg-test-b5 解析失敗

修復 (v3 — 主動創 shared network):
  - B5_NET=b5-test-net (idempotent docker network create)
  - ci-runner 自己 docker network connect $HOSTNAME
  - pg-test-b5 --network=$B5_NET
  - 兩邊同 user-defined network → container name DNS 正常

新增 rule_catalog_sync_job (ADR-090 § Phase 7 第 2 個 service):
  + apps/api/src/jobs/rule_catalog_sync_job.py (~230 行)
    - run_rule_catalog_sync_loop: 啟動延遲 90s,每 1h sync
    - sync_once: HTTP GET {PROMETHEUS_URL}/api/v1/rules (type=alert)
    - UPSERT alert_rule_catalog (rule_name 為 UNIQUE)
    - 只在實際 INSERT/UPDATE 發生時才寫 aol (避免 N 條 rule 污染)
  + main.py lifespan asyncio.create_task() wire

預期解鎖:
  - alert_rule_catalog: 從 0 → Prometheus active rules 數 (~68 條)
  - automation_operation_log: 新增 'rule_created' / 'rule_updated' op_type
  - E3 Hermes AI 終於有 baseline 可以提案規則修正

Refs: ADR-090 §4.2 E3, MASTER §3.3

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-19 16:08:34 +08:00
..