dragonfly-operator icon indicating copy to clipboard operation
dragonfly-operator copied to clipboard
verified publisher icon dragonflydb

Pods are considered ready while loading and cannot serve

Open vyavdoshenko opened this issue 3 months ago • 4 comments

Related: https://github.com/dragonflydb/dragonfly/issues/5881

Problem: Currently, using the standard Docker healthcheck for both liveness and readiness probes causes issues during rolling updates when:

  • New pods pass health checks while still loading datasets into memory
  • Kubernetes kills old pods before new pods are truly ready to serve traffic
  • This results in service downtime during deployments (3+ minutes as mentioned in #5881)

vyavdoshenko avatar Oct 13 '25 11:10 vyavdoshenko

Hi @vyavdoshenko I would like to be assigned on this issue cc @Abhra303

miledxz avatar Nov 10 '25 09:11 miledxz

@miledxz We don't assign issues to comunity authors. You can provide your solution and mention this issue in PR.

vyavdoshenko avatar Nov 10 '25 09:11 vyavdoshenko

@vyavdoshenko sounds good ! thank you for fast response !

miledxz avatar Nov 10 '25 09:11 miledxz

I added a started draft PR here: https://github.com/dragonflydb/dragonfly-operator/pull/412 gentle ping for @Abhra303,

Im open for any feedback, the goal was to do some initial work to see am I on good track

guys thank you for your time !

miledxz avatar Nov 10 '25 11:11 miledxz