ci: 진단 step 에서 모든 backend-spring pod 로그 순회

deployment/<name> selector 는 활성 ReplicaSet 만 봐서 새 ReplicaSet 의
CrashLoopBackOff pod 로그를 놓친다. label selector 로 모든 pod 를
순회하며 current + previous 로그 출력.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.gitea/workflows/deploy.yml

@@ -90,15 +90,21 @@ jobs:
 
           echo
           echo "============================================"
-          echo "=== backend-spring CURRENT logs (tail 400) ==="
+          echo "=== backend-spring ALL pod logs (per pod) ==="
           echo "============================================"
-          kubectl logs -n invyone deployment/backend-spring --tail=400 --all-containers=true || true
-
-          echo
-          echo "============================================"
-          echo "=== backend-spring PREVIOUS logs (crashed pod, tail 400) ==="
-          echo "============================================"
-          kubectl logs -n invyone deployment/backend-spring --tail=400 --previous --all-containers=true || true
+          # deployment/<name> selector 는 active ReplicaSet 한 개만 봐서
+          # 새로 뜨다 죽은 ReplicaSet 의 pod 를 놓침.
+          # label selector 로 모든 backend-spring pod 순회.
+          for p in $(kubectl get pods -n invyone -l app=backend-spring -o name 2>/dev/null); do
+            echo "------------------------------"
+            echo "--- $p (current, tail 500) ---"
+            echo "------------------------------"
+            kubectl logs -n invyone $p --all-containers=true --tail=500 2>&1 || true
+            echo
+            echo "--- $p (previous, if exists, tail 500) ---"
+            kubectl logs -n invyone $p --all-containers=true --tail=500 --previous 2>&1 || true
+            echo
+          done
 
           echo
           echo "============================================"