[
https://issues.apache.org/jira/browse/HBASE-20829?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16529985#comment-16529985
]
Duo Zhang commented on HBASE-20829:
-----------------------------------
The problem is that for MasterProcedureScheduler.yield, the addFront parameter
will be set to false if the proc is not a TableProcedure. But in doAdd method,
if the parent proc has the lock, then it asserts that the addFront parameter
must be true. I think this is a general problem, should be fixed at least for
branch-2.1+. If the pre commit result is OK, will change the title of this
issue to better describe the change, and prepare the patch for branches other
than master.
> TestSyncReplicationStandbyKillRS is flakey
> ------------------------------------------
>
> Key: HBASE-20829
> URL: https://issues.apache.org/jira/browse/HBASE-20829
> Project: HBase
> Issue Type: Bug
> Components: Replication
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Fix For: 3.0.0
>
> Attachments: HBASE-20829-debug.patch, HBASE-20829.patch,
> org.apache.hadoop.hbase.replication.TestSyncReplicationStandbyKillRS-output.txt
>
>
> Timed out.
> {noformat}
> 2018-06-30 01:32:33,823 ERROR [Time-limited test]
> replication.TestSyncReplicationStandbyKillRS(93): Failed to transit standby
> cluster to DOWNGRADE_ACTIVE
> {noformat}
> We failed to transit the state to DA and then wait for it to become DA so
> hang there.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
