canonical / kafka-operator

Kafka VM operator
Apache License 2.0
6 stars 12 forks source link

Removing relation and re-relating with ZooKeeper does not work #218

Open deusebio opened 1 month ago

deusebio commented 1 month ago

When deploying Kafka and Zookeeper, removing the relation and re-adding the relation, not all the Kafka nodes come up healthy. Analysis of the logs suggest that non-leader are not able to authenticate to zookeeper. Indeed the zookeeper-jaas.cfg file on non-leader nodes keep having the old username and password.

We provide a bundle to reproduced attached to the issue.

Steps to reproduce

  1. juju deploy bundle.yaml
  2. (wait for everything to go into a idle/active state)
  3. juju remove-relation kafka zookeeper
  4. juju relate kafka zookeeper

Expected behavior

The kafka cluster comes back online healthy, with all nodes up and running.

Actual behavior

The non leader units report

snap service not running

Versions

Operating system: 22.04

Juju CLI: 3.4/stable

Juju agent: 3.4.4

Charm revision: Kafka 177 and ZooKeeper 134

LXD: 5.0.3

bundle.yaml.txt

github-actions[bot] commented 1 month ago

https://warthogs.atlassian.net/browse/DPE-4984