[HDDS-10875] XceiverRatisServer#getRaftPeersInPipeline should be called before XceiverRatisServer#removeGroup

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.4.0
• Fix Version/s: 2.0.0, 1.4.1
• Component/s: DN
• Labels: pull-request-available
• Target Version/s: 1.4.1
• Assignee: Ivan Andika
• Reporter: Ivan Andika
• Created: 18/May/24 02:20
• Updated: 24/Jul/24 06:40
• Resolved: 21/May/24 03:53

Description

From the https://issues.apache.org/jira/browse/HDDS-10750?focusedCommentId=17847435&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17847435 in HDDS-10750, it's found GroupMismatchException are thrown during the ClosePipelineCommandHandler

This is because XceiverRatisServer#removeGroup is called before XceiverRatisServer#getRaftPeersInPipeline, which causes XceiverRatisServer#getRaftPeersInPipeline to throw GroupMismatchException when it's trying to get the RaftServerProxy#getDivision since the group has been removed.

Therefore, we need to first call the XceiverRatisServer#getRaftPeersInPipeline before calling XceiverRatisServer#removeGroup. On top of that, we can catch the GroupMismatchException in case the group has been removed by earlier ClosePipelineCommandHandler in other datanode for the same pipeline.

 

Issue Links

• is caused by: HDDS-9959 Propagate group remove to other datanodes during pipeline close
• is related to: HDDS-10750 Intermittent fork timeout while stopping Ratis server
• links to: GitHub Pull Request #6696