Scaling Answer replicas in Kubernetes not working

Viewed 110

I have Answer deployed in a cluster currently via the Helm charts included in the repo. The charts have an HPA as a configurable option, but the PVC by default is RWO. While testing out both switching to RWX and limiting replicas to a single host, even scaling the deployment manually up to just two replicas immediately makes sessions freak out. Basically every single action results in a login prompt, until the replicas are scaled back down to 1.

Is scaling Answer replicas, either manually or via the included HPA configurations in the charts actually supported and expected to work?

1 Answers

Maybe we should set the PVC to ReadWriteMany to make it work.

Thanks for the feedback. We probably haven't tested helm enough, so if you find any problems, submit a PR and let's fix them together.