[ https://issues.apache.org/jira/browse/HBASE-25631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17324978#comment-17324978 ]
Xiaolin Ha commented on HBASE-25631: ------------------------------------ There is a dead lock between AsyncFSWAL#waitForSafePoint() and AsyncFSWAL#consume() when calling AsyncFSWAL#doShutdown(). Only consume() can signal condition of `readyForRollingCond`. This is the jstack logs when the RS can not exit and holds for hours. [^jst] > Non-daemon threads make regionserver cannot exit completely > ------------------------------------------------------------ > > Key: HBASE-25631 > URL: https://issues.apache.org/jira/browse/HBASE-25631 > Project: HBase > Issue Type: Bug > Components: regionserver > Affects Versions: 2.0.0, 1.4.7 > Environment: > > Reporter: Xiaolin Ha > Assignee: Xiaolin Ha > Priority: Major > Attachments: 1614845335795-image.png, 1614857532776-image.png, jst > > > When regionserver abort by some errors, the process cannot exit completely. > Error logs and jstack of regionserver process are as follows, > !1614857532776-image.png|width=823,height=364! > * !1614845335795-image.png|width=532,height=330! -- This message was sent by Atlassian Jira (v8.3.4#803005)