From ebf93db1910da1fa932f66327bfca66db3d2298c Mon Sep 17 00:00:00 2001 From: Alert Fix Agent Date: Sun, 26 Jul 2026 05:44:41 +0000 Subject: [PATCH] fix: [alert] KubePodCrashLooping (apps) (#408) --- .../external-dns-unifi/helmrelease.yaml | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/kubernetes/infrastructure/external-dns-unifi/helmrelease.yaml b/kubernetes/infrastructure/external-dns-unifi/helmrelease.yaml index f6fa0836..5b356471 100644 --- a/kubernetes/infrastructure/external-dns-unifi/helmrelease.yaml +++ b/kubernetes/infrastructure/external-dns-unifi/helmrelease.yaml @@ -112,4 +112,13 @@ spec: txtOwnerId: k8s-homelab policy: sync extraArgs: - - --webhook-provider-read-timeout=30s + # The UDM's /proxy/network/v2/.../static-dns/ endpoint occasionally takes + # just over 30s to answer (128 records). external-dns treats ANY RunOnce + # error as fatal (log.Fatalf -> exit 1), so a single slow read crashes the + # container and the pod enters CrashLoopBackOff — even though the webhook + # sidecar completed the fetch successfully a moment later and only failed + # to write the response ("broken pipe", client already gone). 60s gives + # ~2x headroom over the observed worst case. Raising this cannot stall the + # reconcile loop: external-dns runs RunOnce sequentially, so a slow read + # just delays the next cycle past the 1m interval. + - --webhook-provider-read-timeout=60s -- 2.54.0