Open MattySimpLush opened 1 week ago
It is not clear from the issue how do you provision nodeslist to the nodes. Could you add more details about it?
Hi @b0g3r
Yes, so it is deployed as a configmap. For the 3 node setup it looked like below
apiVersion: v1
data:
nodes: |
typesense-0.ts.typesense-search.svc.cluster.local:8107:8108,
typesense-1.ts.typesense-search.svc.cluster.local:8107:8108,
typesense-2.ts.typesense-search.svc.cluster.local:8107:8108
kind: ConfigMap
metadata:
name: nodeslist
namespace: typesense-search
Description
In a Kubernetes-deployed cluster, when one node goes down, the remaining nodes occasionally remove all peers and terminate abruptly. This behaviour disrupts the cluster’s stability and availability.
Steps to reproduce
while true; do; kubectl delete pods -n typesense typesense-0; done
Expected Behavior
When one node in a 3-replica cluster is lost:
Actual Behavior
Example Logs
Metadata
Typesense Version: Observed on 0.25.2 and 27.1. These are the only 2 version I have checked
OS: Running on GKE cos-113-18244-151-88