[
https://issues.apache.org/jira/browse/HBASE-14608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14962934#comment-14962934
]
Heng Chen commented on HBASE-14608:
-----------------------------------
{quote}
if we add throw e just like you mention， sync abort exception will still
interrupt this testcase
{quote}
{code}
if (msg != null && msg.toLowerCase().contains("sync aborted") && i > 50)
{ return; }
+ throw re;
}
{code}
Just as i said above, i think we should NOT add {{throw e}}.
Otherwise it will break down this testcase, we'd better just catch "Sync
aborted exception".
how do you think about it? [~stack]
> testWalRollOnLowReplication has some risk to assert failed after HBASE-14600
> ----------------------------------------------------------------------------
>
> Key: HBASE-14608
> URL: https://issues.apache.org/jira/browse/HBASE-14608
> Project: HBase
> Issue Type: Bug
> Reporter: Heng Chen
> Assignee: Heng Chen
> Fix For: 2.0.0, 1.2.0, 1.3.0
>
> Attachments: HBASE-14608.patch, HBASE-14608_v1.patch
>
>
> After HBASE-14600, we catch runtime exception if dn recover slowly, but it
> has some risk to assert failed.
> For example, https://builds.apache.org/job/HBase-TRUNK/6907/testReport/
> The reason is we catch the exception, but in {{WALProcedureStore}}, it will
> still stop the Procedure. So when we assert stop.isRunning, it will failed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
