Run multinode elasticsearch on K3s

Question

I have installed Elastic/Kibana/Logstash using official helm charts with customized values.yaml on a K3s cluster. If I run kubectl get nodes, I get a list of the cluster nodes correctly. However, when I run kubectl get pods -o, I see all the pods are assigned to only one of the nodes and the remaining nodes are not utilized.

I have tried ➜ ~ kubectl scale --replicas=2 statefulset elasticsearch-master It attempts to schedule the new pods on the same node and triggers pod anti/affinity.

The number of nodes on Kibana stack monitoring is always only 1. The storage is also limited to the first node ephemeral disk.

Should I label the unused cluster nodes explicitly before elastic can start using them?

Could you share the link to the elasticsearc helm chart, the version of the chart and your values.yml file? — antaxify
@antaxify, thank you for your comment, I am using 7.10.2 and I found the error. The error was giving a label to the other nodes on the cluster, I should leave the nodes unlabeled at all. $ kubectl label node ip-X-X-X-X.ec2.internal node-role.kubernetes.io/worker=worker — Mustafa Qamaruddin

Mustafa Qamaruddin Mustafa Qamaruddin · Accepted Answer · 2021-03-13T10:45:53

I found the error. The error was giving a label to the other nodes on the cluster, I should leave the nodes unlabeled at all.

I shouldn't have run:

$ kubectl label node ip-X-X-X-X.ec2.internal node-role.kubernetes.io/worker=worker

Run multinode elasticsearch on K3s

1 Answers