diff --git a/helm/forgejo-db-cluster.yaml b/helm/forgejo-db-cluster.yaml
index a02aab9..db4b474 100644
--- a/helm/forgejo-db-cluster.yaml
+++ b/helm/forgejo-db-cluster.yaml
@@ -22,6 +22,26 @@ metadata:
 spec:
   instances: 1
   imageName: ghcr.io/cloudnative-pg/postgresql:16
+  # Burstable QoS + headroom for pg_dump / forgejo-backup (was BestEffort → probe
+  # timeouts on :8000 under node CPU pressure). Sized like net-kingdom-pg on ~4Gi node.
+  resources:
+    requests:
+      cpu: 200m
+      memory: 512Mi
+    limits:
+      cpu: "1"
+      memory: 1Gi
+  # Default liveness window is 30s (3×10s); extend so backup I/O stalls do not restart the pod.
+  livenessProbeTimeout: 90
+  probes:
+    liveness:
+      timeoutSeconds: 10
+      periodSeconds: 15
+      failureThreshold: 6
+    readiness:
+      timeoutSeconds: 10
+      periodSeconds: 15
+      failureThreshold: 6
   storage:
     size: 10Gi
   bootstrap: