Ethan Rose

I just had some minor comments and questions posted here and on HDDS-7577. I probably won't have time to review this PR, at least not any time soon.

It turns out the pending delete block count can remain high enough on empty containers where it causes starvation in the top N container choosing policy and freezes all pending...

We should make sure this fix works well with [HDDS-7259](https://issues.apache.org/jira/browse/HDDS-7259) and [HDDS-7302](https://issues.apache.org/jira/browse/HDDS-7302) which are newer than this patch, but still don't seem to completely address the problem.

This issue was fixed in #4324 and the jira is resolved. I think we can close this PR. Please reopen if this is not correct.

Thanks for the updates @xBis7. Current changes LGTM, can you just check if the test failures are related? > If we change the ClientVersion to use the latest, someone would...

Yup looks good. Datanode failure should not be related to this change. Key deleting test failure has failed in the same way on [master](https://github.com/elek/ozone-build-results/blob/master/2022/10/27/18074/unit/hadoop-ozone/ozone-manager/org.apache.hadoop.ozone.om.service.TestKeyDeletingService.txt) before. Let me retrigger the tests...

I'll mark these as flaky under HDDS-5626 once we get done here.

> 1. replicationType - RATIS > 2. creationTime - empty string and filter will not be applied, so list out keys irrespective of age, else list out keys which got...

I think just providing the last key in the list as a start key might be easier to resume pagination from than adding a count based offset.

Hi @Montura apologies for the delay. It's late night here currently but I will check this out tomorrow. Thanks for outlining the options.