Problem
--reset-peers-on-error is currently hardcoded in the StatefulSet template (line 49 of templates/statefulset.yaml). This flag causes Typesense to reset all peer connections whenever a peer error is encountered.
During normal catch-up after a pod restart or node join, transient peer errors are common. With this flag always enabled, those errors trigger peer resets which cause additional Raft membership churn, compounding instability in an already-recovering cluster.
Proposed Change
Make the flag opt-in via a values key, e.g.:
typesense:
resetPeersOnError: true # default: true to preserve current behaviour
In the template:
{{- if .Values.typesense.resetPeersOnError }}
- "--reset-peers-on-error"
{{- end }}Why
Operators running stable, well-provisioned clusters should be able to disable this to reduce unnecessary peer resets during catch-up, without having to fork the chart.
Problem
--reset-peers-on-erroris currently hardcoded in the StatefulSet template (line 49 oftemplates/statefulset.yaml). This flag causes Typesense to reset all peer connections whenever a peer error is encountered.During normal catch-up after a pod restart or node join, transient peer errors are common. With this flag always enabled, those errors trigger peer resets which cause additional Raft membership churn, compounding instability in an already-recovering cluster.
Proposed Change
Make the flag opt-in via a values key, e.g.:
In the template:
Why
Operators running stable, well-provisioned clusters should be able to disable this to reduce unnecessary peer resets during catch-up, without having to fork the chart.