Should brokers restart after increasing EBS volume size on their data disks?

[jslusher]

I had the following set up for our 9-broker Kafka cluster's storage:

    storage:
      type: jbod
      volumes:
      - id: 0
        type: persistent-claim
        size: 1500Gi
        deleteClaim: true

I changed size to 2000Gi and I saw that the operator reconciled the podset. At this time 3 of the 9 brokers restarted. They restarted one at a time and everything seemed to go fine, but then that was it. The other 6 brokers did not restart. Execing into the pod and using df -h I can see that both the brokers that restarted and the ones that did not restart have the new volume size.

Why did only 3 of the 9 brokers restart?
Is it necessary at all for the brokers to restart after increasing storage?

[scholzj]

The broker is restarted if the operator sees the PVC in the FileSystemResizePending state. It used to be that this indicated that a restart is needed. Unfortunately, these days it seems it is fifty-fifty, whether it means anything or not, as more and more platforms now support online resizing. If the Pods that did not restart report the new size, you should be fine.