HBASE-28601 Enable setting memstore on-heap sizes in bytes

[liuxiaocs7]

• see: HBASE-28601

[ndimiduk]

LGTM!

[liuxiaocs7]

Hi, @ndimiduk, in the runner summary(https://github.com/apache/hbase/actions/runs/21860359593?pr=7726), how can we see which ut failed, thanks!

[ndimiduk]

The failed units test should be mentioned in a table in the Summary of the run. It's kinda hard to find on the GitHub UI, and it appears to not render at all on mobile.

I'll get you a direct link when I'm back at my desk.

[ndimiduk]

It's this feature. https://github.blog/news-insights/product-news/supercharging-github-actions-with-job-summaries/

[ndimiduk]

I thought I had added this feature along with the unit test job, but I don't see the failed tests in the summary. For now, you'll have to download the artifact from that medium test run and look for the surefire reports there. Let me see what happened to that feature in our build.

[ndimiduk]

I see. I'm parsing Failures but not Errors -- i.e., unit test assertions that failed, but not uncaught thrown exceptions and the like. That's an omission.

[liuxiaocs7]

I see. I'm parsing Failures but not Errors -- i.e., unit test assertions that failed, but not uncaught thrown exceptions and the like. That's an omission.

Hi, @ndimiduk, thank you very much for your quick response. I'm relieved we've identified the underlying issue, which explains why I could sometimes see failed UTs in the Summary and sometimes not. In any case, thank you for your efforts!!!

[liuxiaocs7]

For example, in this runner job summary(https://github.com/apache/hbase/actions/runs/21894809023), we can see Failed Tests in large-wave-3 summary, but nothing in failed large-wave-1 summary.

[ndimiduk]

https://issues.apache.org/jira/browse/HBASE-29888
#7735

[ndimiduk]

The summary I'm talking about is accessed from up here

[liuxiaocs7]

https://issues.apache.org/jira/browse/HBASE-29888

@ndimiduk Got it, thanks Nick!

[liuxiaocs7]

Hi, @ndimiduk, thanks for help to retrigger the ut, medium test fails seems hbase-balancer upgrade to junit5

the first three uts below fails on my local env, maybe they need to upgrade to junit5 because they extends StochasticBalancerTestBase which located in hbase-balancer module

java.lang.NullPointerException: Cannot invoke "org.apache.hadoop.conf.Configuration.setBoolean(String, boolean)" because "org.apache.hadoop.hbase.master.balancer.TestBalancerDecision.conf" is null

	at org.apache.hadoop.hbase.master.balancer.TestBalancerDecision.testBalancerDecisions(TestBalancerDecision.java:60)

Module	Failed Tests
hbase-server	org.apache.hadoop.hbase.master.balancer.TestBalancerDecision.testBalancerDecisions
hbase-server	org.apache.hadoop.hbase.master.balancer.TestBalancerRejection.testBalancerRejections
hbase-server	org.apache.hadoop.hbase.master.balancer.TestCacheAwareLoadBalancerCostFunctions.null
hbase-server	org.apache.hadoop.hbase.replication.regionserver.TestReplicationSource.testAbortTrueOnError

[liuxiaocs7]

Let's see #7744

[liuxiaocs7]

Let's see #7744

With #7744, medium test could pass(https://github.com/apache/hbase/actions/runs/21958775498?pr=7744)!

[liuxiaocs7]

Hi, @ndimiduk, I have rebased master and re trigger ci, failed tests seems unrelated, could you help rerun the failed jobs when free, thanks!

[DieterDePaepe]

Hi, @ndimiduk, I have rebased master and re trigger ci, failed tests seems unrelated, could you help rerun the failed jobs when free, thanks!

I restarted the tests.

[liuxiaocs7]

Hi, @ndimiduk, I have rebased master and re trigger ci, failed tests seems unrelated, could you help rerun the failed jobs when free, thanks!

I restarted the tests.

Thanks for the help, TestSecureIPC fails in large-wave-3 and it should fixed by HBASE-29919(#7794), I'll rebase master and rerun! cc @ndimiduk

[liuxiaocs7]

Hi, @ndimiduk and @DieterDePaepe, could you help re-trigger the failed large-wave-1 and large-wave-3 only, seems the failed ut is unrelated, thanks!!!

[liuxiaocs7]

Hi, Nick, @ndimiduk, is this PR ready to merge now, seems failed ut are all unrelated, thanks!

[ndimiduk]

Hey boss. So sorry for the delay. I was out and no one picked this up. I'll take a look in a few. In the mean time, please check whether we need backports and whether the patch will cherry-pick cleanly.

[liuxiaocs7]

Hey boss. So sorry for the delay. I was out and no one picked this up. I'll take a look in a few. In the mean time, please check whether we need backports and whether the patch will cherry-pick cleanly.

I'm very sorry to bother you. Thank you very much for your suggestion. I will continue to track whether there are conflicts and whether it can be directly backported to other active branches.

[ndimiduk]

Sorry @liuxiaocs7 , I noticed a couple other things.

[liuxiaocs7]

Hi, @PDavid, seems Yetus General Check fails during refguide generating, did I do something wrong here? Could you help take a look when free, thanks!

[PDavid]

Hi, @PDavid, seems Yetus General Check fails during refguide generating, did I do something wrong here? Could you help take a look when free, thanks!

Hi @liuxiaocs7,
I checked the Yetus General Check artifact (yetus-general-check-output archive), inside the patch-refguide.log file. It says the refguide build was successful:

[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  6.028 s
[INFO] Finished at: 2026-03-06T03:00:41Z
[INFO] ------------------------------------------------------------------------

However the log still contains these Maven warnings which I think Yetus considers as a problem:

[WARNING] An issue has occurred with maven-project-info-reports-plugin:3.9.0:dependency-info report, skipping LinkageError 'void org.apache.maven.doxia.sink.Sink.verbatim()', please report an issue to Maven dev team.
java.lang.NoSuchMethodError: 'void org.apache.maven.doxia.sink.Sink.verbatim()'
    at org.apache.maven.reporting.AbstractMavenReportRenderer.verbatimText (AbstractMavenReportRenderer.java:346)
    at org.apache.maven.report.projectinfo.DependencyInformationReport$DependencyInformationRenderer.renderDependencyInfo (DependencyInformationReport.java:218)
    at org.apache.maven.report.projectinfo.DependencyInformationReport$DependencyInformationRenderer.renderMavenDependencyCoordinates (DependencyInformationReport.java:171)

Sorry for this, this was introduced by the changes to the new documentation UI.
I just merged a fix for these #7853 to master.

Can you please try to rebase your branch to have this fix and see if that solves the issue?

[liuxiaocs7]

Hi, @PDavid, many thanks, updated, let's wait the CI!

[ndimiduk]

Thanks @liuxiaocs7 !

[PDavid]

@liuxiaocs7 as I see the refguide check failed again in the Yetus General Check:

  cd /home/runner/work/hbase/hbase/src
  /opt/maven/bin/mvn --batch-mode -Dmaven.repo.local=/home/runner/yetus-m2/hbase-master-patch-608 clean site --batch-mode -pl . -Dtest=NoUnitTests -DHBasePatchProcess -Prelease -Dmaven.javadoc.skip=true -Dcheckstyle.skip=true -Dspotbugs.skip=true > /home/runner/work/hbase/hbase/yetus-general-check/output/patch-refguide.log 2>&1
  [Fri Mar  6 09:50:32 AM UTC 2026 DEBUG]: add_vote_table_v2 >-1< >refguide< >< >10< patch failed to produce the html version of the reference guide.

The refguide is built quite differently now than before and our Yetus check seems to expect the old way:

https://github.com/apache/hbase/blob/master/dev-support/hbase-personality.sh#L456

I think you can ignore it for now until I am able to fix this. Sorry for the hassle.

[liuxiaocs7]

@liuxiaocs7 as I see the refguide check failed again in the Yetus General Check:

  cd /home/runner/work/hbase/hbase/src
  /opt/maven/bin/mvn --batch-mode -Dmaven.repo.local=/home/runner/yetus-m2/hbase-master-patch-608 clean site --batch-mode -pl . -Dtest=NoUnitTests -DHBasePatchProcess -Prelease -Dmaven.javadoc.skip=true -Dcheckstyle.skip=true -Dspotbugs.skip=true > /home/runner/work/hbase/hbase/yetus-general-check/output/patch-refguide.log 2>&1
  [Fri Mar  6 09:50:32 AM UTC 2026 DEBUG]: add_vote_table_v2 >-1< >refguide< >< >10< patch failed to produce the html version of the reference guide.

The refguide is built quite differently now than before and our Yetus check seems to expect the old way:

https://github.com/apache/hbase/blob/master/dev-support/hbase-personality.sh#L456

I think you can ignore it for now until I am able to fix this. Sorry for the hassle.

@PDavid, got it, many thanks for your feedback! And I've verified locally that both the generated web pages and the refguide works well!