[ https://issues.apache.org/jira/browse/HBASE-20651?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16524605#comment-16524605 ]
huaxiang sun commented on HBASE-20651: -------------------------------------- The issue is that hbck sends a offline command to offline a split parent region, which will change region's in-memory state from SPLIT to OFFLINE, and a following assign() will succeed and bring back the SPLIT parent region to life. I added a simple check, though it wont address the issue totally (due to no lock protection to change region's in-memory state), the sanity check will filter out most of corner cases (hopefully) > Master, prevents hbck or shell command to reassign the split parent region > -------------------------------------------------------------------------- > > Key: HBASE-20651 > URL: https://issues.apache.org/jira/browse/HBASE-20651 > Project: HBase > Issue Type: Improvement > Components: master > Affects Versions: 1.2.6 > Reporter: huaxiang sun > Assignee: huaxiang sun > Priority: Minor > Attachments: HBASE-20651-branch-1-v001.patch > > > We are seeing that hbck brings back split parent region and this causes > region inconsistency. More details will be filled as reproduce is still > ongoing. Might need to do something at hbck or master to prevent this from > happening. -- This message was sent by Atlassian JIRA (v7.6.3#76005)