d5a4e273449541b47d13ba1d4b0af1d73551afbe
Some checks failed
CD Pipeline / deploy (push) Has been cancelled
問題:
Phase 38 加了 host_health_probes 表 + 開觀測台頁面時寫一筆,但
無人開頁時沒人寫 → Telegram cmd:obs_health 顯示「24h uptime」永遠空。
修補:
- run_scheduler.py::run_host_health_probe
- 每 15 min HTTP probe GCP-A/GCP-B/111 三主機 /api/tags
- 寫入 host_health_probes(label/url/healthy/unhealthy_mark/
models_count/response_ms/error_msg)
- 失敗安全:HTTP/DB 失敗只 log warning
- run_scheduler.py::run_host_health_probe_cleanup
- 每日 03:00 DELETE 30d 前舊資料(防表膨脹)
- 註冊到 schedule.every(15).minutes 與 schedule.every().day.at("03:00")
效果:
- Web /observability/host_health 24h 趨勢卡永遠有資料(即使無人開頁)
- Telegram cmd:obs_health 三主機在線率永遠有資料
- 三主機歷史完整保留 30 天,超出自動清理
Phase 38+39+40+41+42 觀測台戰役完整收官(7 commits)。
部署驗證:
- mo.wooo.work/observability/host_health → HTTP 200 / 42716 byte
(Phase 38 為 39124 byte,多 3.5KB 證明 24h 趨勢/MCP/AIOps card 已上線)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Description
EwoooC — 商品看板 + 業績報表 + AI KM (Flask + pgvector, Docker Compose on 188)
Languages
PostScript
59.7%
Python
30.9%
HTML
4.2%
CSS
2.1%
JavaScript
1.9%
Other
1.1%