Hello,
I solved my problem by changing livenessProbe and readynessProbe in Kubernetes.
Jira status page during re-indexation sends a 503 HTTP code, so Kubernetes considers the application in failure and it restarts the pod before the reindexing finishes.
I now watch out the TCP port and not the HTTP code on /status page and it works.
I think that Jira shouldn’t send a 500 HTTP code while re-indexing because it is a “normal” and “wanted” operation launched by an administrator of the application.
For the moment, I edited Helm Chart that I used (https://github.com/stevehipwell/helm-charts/pull/232).
Am I the only person who faced this problem ?