IOPS was too high when mq was cleanning up expired files automatically

[H3nryMann]

Search before creation

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

Documentation Related

Cluster's flush type is SYNC;
Cluster's deployment type is 2m-2s;
Cluster's version is 4.9.7;
300-500 topics in cluster;
And other runtime settings are almost default.
I found the IOPS was too high when deleted expired files automatically in CommitLog, IndexFile and ConsumeQueue.
What reasons could lead to this? How to decrease IO.

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[ChangingFond]

You can increase these two configuration values. The default is 100ms.

deleteCommitLogFilesInterval
deleteConsumeQueueFilesInterval

[lizhimins]

It appears you are using a file-based CQ implementation.
In version 5.x, binary search is used to reduce the I/O of the correct min CQ offset process.