[
https://issues.apache.org/jira/browse/HBASE-20777?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16521313#comment-16521313
]
Duo Zhang commented on HBASE-20777:
-----------------------------------
https://builds.apache.org/job/HBASE-Flaky-Tests/33593/testReport/junit/org.apache.hadoop.hbase.replication/TestReplicationKillMasterRS/killOneMasterRS/
Seems the same problem. Connecting to an aborted RS leads to connection timed
out, not connection refused. So the problem maybe that the abort method for RS
does not work correctly. Let me dig more.
> TestAsyncTableBatch is flakey
> -----------------------------
>
> Key: HBASE-20777
> URL: https://issues.apache.org/jira/browse/HBASE-20777
> Project: HBase
> Issue Type: Bug
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Attachments: HBASE-20777.patch,
> org.apache.hadoop.hbase.client.TestAsyncTableBatch-output.txt
>
>
> The log is very strange, we keep sending request to a dead RS, and the result
> is not connection refused, but rpc timeout, and later it becomes
> CallQueueTooBig...
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
