fix(runner): retry harbor repair ssh probe

.gitea/workflows/harbor-110-local-repair.yaml

@@ -65,9 +65,26 @@ jobs:
             -o ServerAliveCountMax=2
             "${AWOOOI_110_SSH_TARGET}"
           )
+          SSH_PROBE_ATTEMPTS="${AWOOOI_110_SSH_PROBE_ATTEMPTS:-6}"
+          SSH_PROBE_SLEEP_SECONDS="${AWOOOI_110_SSH_PROBE_SLEEP_SECONDS:-10}"
 
           run_ssh() {
-            timeout 30 "${ssh_base[@]}" "$@"
+            local attempt rc
+            attempt=1
+            rc=1
+            while [ "${attempt}" -le "${SSH_PROBE_ATTEMPTS}" ]; do
+              if timeout 30 "${ssh_base[@]}" "$@"; then
+                echo "harbor_110_remote_ssh_probe_attempt=${attempt} result=success"
+                return 0
+              fi
+              rc=$?
+              echo "harbor_110_remote_ssh_probe_attempt=${attempt} result=failure rc=${rc}"
+              if [ "${attempt}" -lt "${SSH_PROBE_ATTEMPTS}" ]; then
+                sleep "${SSH_PROBE_SLEEP_SECONDS}"
+              fi
+              attempt=$((attempt + 1))
+            done
+            return "${rc}"
           }
 
           diagnose_ssh_control_channel() {

docs/LOGBOOK.md

@@ -1,3 +1,17 @@
+## 2026-07-01 — 09:44 Harbor repair SSH probe bounded retry
+
+**照主線修正的問題**：
+- 最新 live truth：CD `#4215` 仍因 Harbor public `/v2/` = `502` 失敗；Harbor repair `#4212` 的具體 blocker 是 `harbor_110_remote_control_channel_unavailable`。
+- 188 non-110 runner lane 讀回 ready、host pressure 正常；但 188 → 110 bounded SSH probe 呈現間歇性，一次 `true` 可成功，下一次 `recover-110-control-path-and-harbor-local.sh --check` 又 timeout。
+- `.gitea/workflows/harbor-110-local-repair.yaml` 對非寫入的 SSH probe / verifier 加 bounded retry：預設 `6` 次、每次仍受 `ConnectTimeout=8`、`ServerAlive*` 與外層 `timeout 30` 限制，並輸出 `harbor_110_remote_ssh_probe_attempt=...` receipt。`run_recovery --apply-all` 不自動 retry，避免半套用被重跑。
+
+**驗證**：
+- `python3.11 -m pytest ops/runner/test_cd_controlled_runtime_profile.py -q`：`35 passed`。
+- `DATABASE_URL=postgresql+asyncpg://test:test@localhost:5432/test PYTHONPATH=apps/api python3.11 -m pytest ops/runner/test_read_public_gitea_actions_queue.py ops/runner/test_cd_controlled_runtime_profile.py ops/runner/test_install_awoooi_non110_runner_user_service.py ops/runner/test_verify_awoooi_non110_cd_closure.py apps/api/tests/test_harbor_registry_controlled_recovery_receipt.py -q`：`96 passed`。
+- `python3.11 ops/runner/guard-gitea-runner-pressure.py --root .`、`node scripts/ci/check-gitea-step-env-secrets.js .gitea/workflows/harbor-110-local-repair.yaml`、YAML parse、`git diff --check`：通過。
+
+**邊界**：只改 Harbor repair workflow 的 bounded non-write SSH probe retry / tests / LOGBOOK；未讀 secret / token / `.env` / raw sessions / SQLite / auth；未讀 `.runner` 內容；未使用 GitHub / `gh` / GitHub API；未 workflow_dispatch；未重啟主機、未 restart Docker / Nginx / K3s / DB / firewall。
+
 ## 2026-07-01 — 09:36 188 runner drain / 110 stale docker CPU metric correction
 
 **照主線修正的問題**：

ops/runner/test_cd_controlled_runtime_profile.py

@@ -136,6 +136,12 @@ def test_harbor_110_local_repair_workflow_is_dispatch_only_and_bounded() -> None
     assert "AWOOOI_110_SSH_TARGET" in text
     assert "BatchMode=yes" in text
     assert "ConnectTimeout=8" in text
+    assert 'SSH_PROBE_ATTEMPTS="${AWOOOI_110_SSH_PROBE_ATTEMPTS:-6}"' in text
+    assert (
+        'SSH_PROBE_SLEEP_SECONDS="${AWOOOI_110_SSH_PROBE_SLEEP_SECONDS:-10}"'
+        in text
+    )
+    assert "harbor_110_remote_ssh_probe_attempt=" in text
     assert "operation_boundary_remote_ssh_bounded=true" in text
     assert "harbor_110_remote_control_channel_unavailable" in text
     assert "harbor_110_remote_repair_check_start=1" in text