[ https://issues.apache.org/jira/browse/CASSANDRA-18905?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17776792#comment-17776792 ]
Andres de la Peña commented on CASSANDRA-18905: ----------------------------------------------- The test failures don't seem caused by the changes: * {{testServiceTopPartitionsSingleTable}} is CASSANDRA-17798. * {{simulationTest}} is not reported on Butler nor JIRA but can be reproduced on [5.0|https://app.circleci.com/pipelines/github/adelapena/cassandra/3253/workflows/e48b49e9-cf36-412a-a811-d813031e6f01/jobs/83735/tests] and [trunk|https://app.circleci.com/pipelines/github/adelapena/cassandra/3254/workflows/69f451ef-fb39-48e4-b1d1-40ee4141b0c1/jobs/83739/tests]. * {{explicitEndpointIgnore}} is not reported on Butler nor JIRA but can be reproduced on [trunk|https://app.circleci.com/pipelines/github/adelapena/cassandra/3254/workflows/69f451ef-fb39-48e4-b1d1-40ee4141b0c1/jobs/83738/tests]. I'll open tickets for the two unreported test failures. > Index.Group is incorrectly unregistered from the SecondaryIndexManager > ---------------------------------------------------------------------- > > Key: CASSANDRA-18905 > URL: https://issues.apache.org/jira/browse/CASSANDRA-18905 > Project: Cassandra > Issue Type: Bug > Components: Feature/2i Index > Reporter: Mike Adamson > Assignee: Mike Adamson > Priority: Urgent > Fix For: 5.0, 5.1 > > > An Index.Group is removed from the SecondaryIndexManager during > unregisterIndex if it contains no indexes after the index is unregistered. > The code for removing the group uses the wrong key to remove the group from > the indexGroups map. It is using the group object rather than the group name > that is used as the key in the map. > This means that the group is not added again if a new index is registered > using that group. The knock on from this is that the > StorageAttachedIndexGroup unregisters itself from the Tracker when it has no > indexes after an index is removed. The same group with no tracker is then > used for new indexes. This group then receives no notifications about sstable > or memtable updates. The ultimate side effect of this is that, memtables are > not released, resulting in memory leaks and indexes are not updated with new > sstables and their associated index files. -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org