Problem Statement

rpk cluster health reports a leaderless partition but the partition actually reports a leader according to rpk topic describe -p . The following example output of rpk topic describe denotes the leader is elected for partition 0.

PARTITIONS
==========
PARTITION  LEADER  EPOCH  REPLICAS  LOG-START-OFFSET  HIGH-WATERMARK
0          2       79     [0 1 2]   0                 13

This is a bug in Redpanda where the rpk cluster health reports stale info about leaderless partitions and applies the following versions:

• v23.3.11 and lower
• v23.2.27 and lower

Fixed versions are:

• v23.3.12 and higher
• v23.2.28 and higher

As the leader is actually elected, this doesn't affect the user application either with producing or consuming. This is a problem in reporting right status of the cluster.

Solution

Upgrade to the fixed version by following the instruction in https://docs.redpanda.com/current/upgrade/.

If upgrade is not a feasible solution for a short term, report it to the Redpanda support team to mitigate the stale info .

 

Notes

Add any references to further reading on the question.