[ 
https://issues.apache.org/jira/browse/HDDS-17501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18133214#comment-18133214
 ] 

Sergey Soldatov commented on HDDS-17501:
----------------------------------------

 Seen in CI: 
https://github.com/ss77892/ozone/actions/runs/40553006606/job/122375507065
 
  
TestBackgroundContainerDataScannerIntegration>ContainerScannerIntegrationTests.shutdown:116
  java.lang.AssertionError: Found 1 metrics source(s) still registered after 
cluster shutdown: [RDBMetrics]
  Root cause: {{OzoneContainer#stopContainerScrub}} returns early when 
{{metadataScanner == null}}. Since HDDS-8553, the metadata scanner is optional
  ({{hdds.container.scrub.metadata.enabled}}). When it is disabled, the 
background data scanners and the on-demand scanner are never shut down. The 
test hits this now because
  HDDS-17301 fixed the config key, so {{setMetadataScanEnabled(false)}} 
actually takes effect.

  How the leak happens:
  {{volumeSet.shutdown()}} closes the per-volume RocksDB. {{RDBStore#close}} 
calls {{RDBMetrics.unRegister()}}, which unregisters the source shared by all 
stores in the JVM.
   A data scanner that is still running calls {{BlockUtils.getDB()}}. That 
reopens the volume DB through {{DatanodeStoreCache}}, which registers 
{{RDBMetrics}} again.
  In mini cluster mode, {{DatanodeStoreCache#shutdownCache}} keeps its entries. 
The reopened DB is never closed, and {{MetricsLeakAssertion}} reports 
{{RDBMetrics}}.

 The failure is intermittent because a scanner has to be scanning a container 
just after the volumes are closed.

> OzoneContainer#stopContainerScrub do not stop all scanners
> ----------------------------------------------------------
>
>                 Key: HDDS-17501
>                 URL: https://issues.apache.org/jira/browse/HDDS-17501
>             Project: Apache Ozone
>          Issue Type: Bug
>            Reporter: Ivan Andika
>            Assignee: Jheng-Sing Chen
>            Priority: Minor
>
> See https://github.com/apache/ozone/pull/11773#discussion_r4379237308



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to