Details
-
Bug
-
Resolution: Fixed
-
Minor
-
1.1.0, 1.2.0
Description
- Created a cluster with 3 pods having Persistent volume storage as Gluster.
- Created 2 buckets with 0 replica (Used the same configuration files provided by you previously)
- Once the three pods were running successfully, deleted one of the pods using oc delete pod <pod name>.
- The auto-failover timeout was set to 10 seconds. Since the replica for the buckets were 0, the auto-failover could not be triggered as expected to prevent loss of data.
- The pod went down and a new pod with the same name came up (due to the existing PVs).
- The Rebalance button is enabled on the UI at this point (as shown in the screenshot above) even though the node is still in the cluster. The writes to this node in Pending mode were affected.
- Once the Rebalance button was manually clicked, all the nodes became green and every node started taking the write operations in the usual way.
- Performed a test on a cluster deployed normally using RPM binaries (not using Couchbase Operator). Observed the same behavior where the Rebalance button was enabled.
- Couchbase Operator was not able to automatically trigger the Rebalance process in this scenario.
Need to further investigate about why the Operator was not able to trigger the Rebalance automatically.
Logs:
https://supportal.couchbase.com/snapshot/c7248c67712e33e096f4a18d0344b931::0