From 8558ac2d20de2d9f463c3d0a682d7453d504c586 Mon Sep 17 00:00:00 2001
From: Your Name <you@example.com>
Date: Sun, 24 May 2026 10:11:20 +0800
Subject: [PATCH] fix(k8s): use lightweight api probes

---
 k8s/awoooi-prod/06-deployment-api.yaml | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/k8s/awoooi-prod/06-deployment-api.yaml b/k8s/awoooi-prod/06-deployment-api.yaml
index e1550ff1..3c3ddfed 100644
--- a/k8s/awoooi-prod/06-deployment-api.yaml
+++ b/k8s/awoooi-prod/06-deployment-api.yaml
@@ -25,7 +25,9 @@ spec:
     type: RollingUpdate
     rollingUpdate:
       maxSurge: 1
-      maxUnavailable: 0
+      # 2026-05-24 Codex: allow one unavailable replica so rollout can replace
+      # a bad old ReplicaSet instead of deadlocking at 1/2 when probes regress.
+      maxUnavailable: 1
   template:
     metadata:
       labels:
@@ -137,7 +139,10 @@ spec:
               memory: "1Gi"
           livenessProbe:
             httpGet:
-              path: /api/v1/health
+              # 2026-05-24 Codex: K8s probes must stay lightweight. The full
+              # health endpoint checks Ollama/OpenClaw/SigNoz and can exceed
+              # kubelet timeout when an external provider is degraded.
+              path: /api/v1/health/live
               port: 8000
             initialDelaySeconds: 30
             periodSeconds: 10
@@ -145,7 +150,7 @@ spec:
             failureThreshold: 3
           readinessProbe:
             httpGet:
-              path: /api/v1/health
+              path: /api/v1/health/ready
               port: 8000
             initialDelaySeconds: 5
             periodSeconds: 5
@@ -154,7 +159,7 @@ spec:
           # Phase K0.5: Startup Probe (允許最多 60 秒啟動時間)
           startupProbe:
             httpGet:
-              path: /api/v1/health
+              path: /api/v1/health/live
               port: 8000
             initialDelaySeconds: 5
             periodSeconds: 5