Merge remote-tracking branch 'gitea-ssh/main' into codex/github-redacted-evidence-validator-20260627
Some checks failed
Ansible / Reboot Recovery Contract / validate (push) Successful in 1m16s
CD Pipeline / tests (push) Successful in 1m42s
Code Review / ai-code-review (push) Successful in 15s
CD Pipeline / build-and-deploy (push) Failing after 15m29s
CD Pipeline / post-deploy-checks (push) Has been cancelled

This commit is contained in:
Your Name
2026-06-28 09:13:32 +08:00
8 changed files with 26 additions and 10 deletions

View File

@@ -289,9 +289,9 @@ force push / 刪 repo / 刪 refs / 改 repo visibility / raw runtime secret volu
### 110 runner 壓力事故例外
2026-06-28 事故後110 上的 Gitea / act-runner、StockPlatform headless smoke、host-side Next build 與 Docker / BuildKit 壓力屬容量事故保護面。即使收到「批准 / 繼續 / 全面授權」,也不得直接重開 runner、解除 service mask、還原 live runner binary、恢復泛用 `ubuntu-latest` label或把 host pressure gate 改成 warn-only 作為預設。
2026-06-28 事故後110 上的 Gitea / act-runner / direct transient runner、StockPlatform headless smoke、host-side Next build 與 Docker / BuildKit 壓力屬容量事故保護面。即使收到「批准 / 繼續 / 全面授權」,也不得直接重開 runner、解除 service mask、還原 live runner binary、用 `systemd-run` 直啟 `.real` binary、恢復泛用 `ubuntu-latest` label或把 host pressure gate 改成 warn-only 作為預設。
允許的 controlled apply 是降壓與防再發:停止 / disable / mask runner、quarantine runner binary、收斂 labels、補 source fail-closed guard、搬遷 runner、限制 concurrency、把 smoke 改成排程 / 非 110 runner以及執行只讀 pressure / cold-start verifier。
允許的 controlled apply 是降壓與防再發:停止 / disable / mask runner、mask direct transient unit、quarantine runner binary、收斂 labels、補 source fail-closed guard、搬遷 runner、限制 concurrency、把 smoke 改成排程 / 非 110 runner以及執行只讀 pressure / cold-start verifier。
恢復 runner 必須同時具備:

View File

@@ -30,6 +30,7 @@
- Live 110 `/usr/local/bin/awoooi-wait-host-web-build-pressure.sh``/usr/local/bin/awoooi-startup-110.sh` 已同步 fail-closed 並加 immutable。
- Live 110 四條 runner 入口改為 immutable fail-closed stub原 ELF 僅 quarantine 不讀內容:`/home/wooo/act-runner/act_runner``/home/wooo/act-runner/act_runner.real-20260628-runner-pressure-guard``/home/wooo/act-runner-controlled/act_runner``/home/wooo/awoooi-controlled-runner/awoooi_controlled_runner`
- System-level 與 user-level `gitea-act-runner-host.service``gitea-act-runner-awoooi-controlled.service``gitea-awoooi-controlled-runner.service``gitea-act-runner-awoooi-open.service` 皆讀回 inactive / masked原 unit file 僅 quarantine並以 `/dev/null` mask symlink 防止直啟。
- 08:54 又抓到 transient `awoooi-direct-runner-open.service` 直啟 `.real` binary且把 `/home/wooo/act-runner/act_runner` / `.real` 還原成 ELF已強制 kill、將 `awoooi-direct-runner-open.service``awoooi-direct-runner.service` 建立 `/dev/null` mask並再次 quarantine ELF / 還原 immutable fail-closed stub。
- `AGENTS.md``docs/HARD_RULES.md``ops/runner/README.md` 與 MASTER runner 章節補上 110 runner 壓力事故例外:全面授權不等於可重開 production host runner。
**Live readback**
@@ -37,6 +38,9 @@
- 08:40 曾抓到 parent=1 的孤兒 `/home/wooo/act-runner/act_runner.real-20260628-runner-pressure-guard daemon --config config.yaml`,已 terminate後續改為 stub 以防任何 direct exec 繞過 systemd。
- 08:47 延遲讀回:四個 system runner unit 皆 `LoadState=masked``ActiveState=inactive``UnitFileState=masked`user-level runner units 亦為 masked精準 runner process scan 無命中。
- 08:52 以乾淨 SSH command 重跑 live pressure gate`/usr/local/bin/awoooi-wait-host-web-build-pressure.sh``GATE_RC=0``no host web/build/smoke pressure detected`
- 08:56 讀回:`awoooi-direct-runner-open.service``awoooi-direct-runner.service` 與四個 Gitea runner units 全部 `masked / inactive`;四條 runner binary 皆為 163-byte immutable shell stub`pgrep` 無 runner processpressure gate 回 `GATE_RC=0`
- 08:58 延遲讀回曾顯示 transient `awoooi-direct-runner-open.service` 已消失成 `not-found`;已立即補 `systemctl mask awoooi-direct-runner-open.service`,確認 `/etc/systemd/system/awoooi-direct-runner-open.service -> /dev/null``LoadState=masked` / `UnitFileState=masked`
- 08:59 最終短延遲讀回:兩個 direct runner units 與四個 Gitea runner units 全部 `masked / inactive`runner process scan 無命中pressure gate 回 `GATE_RC=0`
- `post-start-quick-check.sh --no-color``SCORECARD_RC=2``POST_START_QUICK_CHECK PASS=35 WARN=4 BLOCKED=4``RESULT=BLOCKED`public routes 全部 HTTP OKStockPlatform freshness `ok` / latest trading date `2026-06-26`,但 MOMO daily sales data stale `4|2026-06-24`、backup heartbeat core blocker、credential escrow missing `5` 仍阻擋全主機 green。
- 08:49 曾觀察到 `/home/wooo/awoooi-manual-deploy` 的手動 Web image buildparent 是另一條 SSH session不是 Gitea runner。08:52 乾淨 gate 已恢復 `GATE_RC=0`;此 live-only manual deploy path 仍需後續納入 pressure lock / 非 110 build path。