[
https://issues.apache.org/jira/browse/HBASE-22416?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16839680#comment-16839680
]
Josh Elser commented on HBASE-22416:
------------------------------------
bq. we may still need the -Djava.net.preferIPv4Stack=true to avoid subsequent
test failures.
Today, I couldn't get a successful full run without this. Whereas, with the
patch here, I didn't need to set {{preferIPv4Stack}} at all.
bq. I guess the proposed patch additions on ZKTreeLockManager make sense
Yeah, I think that's still relevant (heavy-handed locking). Just circumstantial
that we see this now :)
> HBOSS: unit tests fail with ConnectionLoss when IPv6 enabled and not set up
> locally
> -----------------------------------------------------------------------------------
>
> Key: HBASE-22416
> URL: https://issues.apache.org/jira/browse/HBASE-22416
> Project: HBase
> Issue Type: Bug
> Reporter: Josh Elser
> Assignee: Josh Elser
> Priority: Minor
> Fix For: hbase-filesystem-1.0.0-alpha1
>
> Attachments: HBASE-22416.001.patch
>
>
> I've seen that the HBOSS unit tests will fail for me with a ConnectionLoss
> exception. This is resolved when I force ipv4. My hunch is that we don't
> block for curator to become connected, and because of my local setup, the
> ipv6 lookup happens first and has to time-out. The test starts running but
> Curator hasn't yet become connected, so we fail with a ConnectionLoss in the
> test itself.
> We should wait for Curator to become connected and maybe addÂ
> {{-Djava.net.preferIPv4Stack=true}} to surefire opts
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
