Search before asking

• I had searched in the issues and found no similar issues.

Version

0.14.13

What's Wrong?

in my doris cluster, tablet frequent balance, a tablet are added to the balance pending queue every second, but most of the balance tasks failed, few tasks can be successful. when success, the redundant replication will be deleted, but the next second will still enter the pending balance queue. When I delete the table to which this tablet belongs, other one tablet will appear in the balance pending queue every second. and this situation has been going on for many days.

I analyzed the cause of the problem. In my doris cluster, each be has two kinds of disk medium, ssd and hdd, and most of the data is used ssd. For historical reasons, there are very few tables in hdd, and these hdd table occupies a very small space, only at the k level, but the hdd disk space is very large, at the T level. When the amount of data is small, the balance strategy is difficult to balance the data between all be, which means difficult to let all be are in the "MID" state. Check SHOW PROC'/cluster_balance/cluster_load_stat/HDD', there are always some bes in the "HIGN" state, and some bes in the "MID" or "LOW" state.

What You Expected?

A tablet, don’t need to be frequently balanced

How to Reproduce?

• No response*

Anything Else?

• No response*

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct