From a0d4bc9de2becc31fbe935e16e25156c9805bbfb Mon Sep 17 00:00:00 2001 From: tegwick Date: Tue, 7 Jul 2026 18:11:26 +0200 Subject: [PATCH] Stabilize forgejo-db CNPG pod under backup load Add Burstable CPU/memory requests and relax liveness/readiness probes so pg_dump and forgejo-backup I/O no longer trip :8000 health checks on the ~4Gi railiance01 node (was BestEffort with 5 probe-driven restarts). --- helm/forgejo-db-cluster.yaml | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) 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: