[
https://issues.apache.org/jira/browse/ZOOKEEPER-1768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13794244#comment-13794244
]
Flavio Junqueira commented on ZOOKEEPER-1768:
---------------------------------------------
If it is what I think it is, then you need to increase the initLimit. I don't
know of a way of changing the initLimit dynamically, though. Also, another
problem seems to be that the autopurge feature is removing correct snapshots
and leaving partial snapshots on disk. If we can't complete the write of the
snapshot, then perhaps we should delete it write after learning it. Given the
behavior you're describing, I believe this is not what we are doing.
> Cluster fails election loop until the device is full
> ----------------------------------------------------
>
> Key: ZOOKEEPER-1768
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1768
> Project: ZooKeeper
> Issue Type: Bug
> Components: leaderElection
> Affects Versions: 3.4.5
> Reporter: yuxin.yan
> Fix For: 3.4.6, 3.5.0
>
> Attachments: zk_debug.log.2013-09-25.log, zoo.cfg
>
>
> Hi,
> I have a five nodes cluster versioned 3.4.5 and now i find one node is
> offline.
> Firstly i restart the node but i find that "Error contacting service. It is
> probably not running." and i find that the node always elect the leader and
> always sync the snapshot logs and the device will be full every ten mins.
> so could someone help me？ i will put the log and zoo.cfg in the attachment.
> Thanks all.
> yyx,
--
This message was sent by Atlassian JIRA
(v6.1#6144)
