[
https://issues.apache.org/jira/browse/PHOENIX-56?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13903520#comment-13903520
]
Lars Hofhansl commented on PHOENIX-56:
--------------------------------------
This is really scary. Especially the "need to restart HBase" part.
We do not have much to go by, though.
It would be good have stack and heap dumps from both the client and one or two
of the regionservers to see what's going on.
Does this look like a memory leak, Russel? Does client and/or regionserver
memory usage grow without bounds?
Is there a way to write a script to reproduce this behavior reliably?
> Phoenix/JDBC becomes unresponsive
> ---------------------------------
>
> Key: PHOENIX-56
> URL: https://issues.apache.org/jira/browse/PHOENIX-56
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 2.2.3
> Environment: HBase 0.94.6
> Reporter: Russell Jurney
>
> After some hours or days, our Phoenix JDBC connection becomes unresponsive.
> It can't be reached unless we remove all tables, including SYSTEM.TABLE and
> restart HBase.
> There are no errors that accompany this problem, so we don't know how to
> debug it. Is anyone else seeing this behavior?
> Using a different client version of the JAR does result in an error saying
> version mismatch... so at some level we think the service is running. It just
> can't complete a connection. sqlline.sh freezes when we connect.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
