stability: harden fluent-bit buffering and longhorn node-down recovery

2026-04-27 16:15:13 -03:00 · 2026-04-27 16:15:13 -03:00 · 9fb8dd4839
commit 9fb8dd4839
parent 6352e0d976
2 changed files with 2 additions and 2 deletions
--- a/infrastructure/longhorn/core/scripts/longhorn_settings_ensure.sh
+++ b/infrastructure/longhorn/core/scripts/longhorn_settings_ensure.sh
@ -44,3 +44,4 @@ update_setting support-bundle-manager-image "registry.bstein.dev/infra/longhorn-
 # Keep storage-heavy nodes from getting hammered by rebuild storms and skew.
 update_setting replica-auto-balance "best-effort"
 update_setting concurrent-replica-rebuild-per-node-limit "2"
+update_setting node-down-pod-deletion-policy "delete-both-statefulset-and-deployment-pod"
--- a/services/logging/fluent-bit-helmrelease.yaml
+++ b/services/logging/fluent-bit-helmrelease.yaml
@ -44,8 +44,7 @@ spec:
          path: /var/log/journal
      - name: fluentbit-state
        emptyDir:
-          medium: Memory
-          sizeLimit: 64Mi
+          sizeLimit: 1Gi
    extraVolumeMounts:
      - name: runlogjournal
        mountPath: /run/log/journal