[
https://issues.apache.org/jira/browse/HBASE-19210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16251729#comment-16251729
]
Mike Drob commented on HBASE-19210:
-----------------------------------
If we expect a socket exception we should check that it was actually thrown. I
don't like that the behaviour is inconsistent depending on previous calls - it
is very un-REST like to me.
Can you try figuring out if it is a client or server caused error? Does getting
a new client make the error go away? Is it a problem with too many server
errors where we end up somehow losing threads?
> TestNamespacesInstanceResource fails
> ------------------------------------
>
> Key: HBASE-19210
> URL: https://issues.apache.org/jira/browse/HBASE-19210
> Project: HBase
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Ted Yu
> Fix For: 2.0.0-beta-1
>
> Attachments: 19210.addendum, 19210.v1.txt, 19210.v1.txt,
> 19210.v2.txt, 19210.v3.txt, HBASE-19210.WIP.patch
>
>
> This is the top flaky test.
> The following can be reproduced:
> {code}
> java.net.SocketException: Connection reset
> at
> org.apache.hadoop.hbase.rest.TestNamespacesInstanceResource.testInvalidNamespacePostsAndPuts(TestNamespacesInstanceResource.java:271)
> {code}
> With commit e320df5a0c267258c03909da8d0eee4c0e287532, the test passes.
> With commit 5facaded902a13556952b1f9d26b768cb86e6599, the test fails.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
