[Bug] Brokers can't reach balanced state sometimes when using UnifromLoadShedder

[aloyszhang]

Search before asking

• I searched in the issues and found nothing similar.

Version

2.10.5

Minimal reproduce step

1. Start 3 brokers and create some topics
2. Start producing and consuming for all these topics
3. Enable load balance with Uniformloadshedder:

• loadBalancerMsgThroughputMultiplierDifferenceShedderThreshold=1.1
• loadBalancerMsgRateDifferenceShedderThreshold=0, means disable msgRate based shedding

5. Use the default loadBalancerLoadPlacementStrategy org.apache.pulsar.broker.loadbalance.impl.LeastLongTermMessageRate
6. Restart brokers several times to trigger the load balance
7. There is a chance to see brokers can't reach the balanced state for long time.

What did you expect to see?

All brokers should reach a balanced state after rounds of shedding.

What did you see instead?

Brokers may stay in an un-balanced state for a long time.

Anything else?

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!

[github-actions]

The issue had no activity for 30 days, mark with Stale label.