[sheepdog] [PATCH v2] zk: actively remove itself after leave_cluster()is called

輕微 cz at de3eb.cn
Mon May 13 10:05:13 CEST 2013


after a node leaved(e.g. #collie cluster shutdown), it can not start up again before zk timeout. 
the node broadcast a EVENT_LEAVE msg, but no body told zookeeper to remove it.


so there will be a log msg like this:
zk_join(500) Previous zookeeper session exist, shoot myself.
until zk timeout.
 
 
------------------ Original ------------------
From:  "Liu Yuan"<namei.unix at gmail.com>;
Date:  Mon, May 13, 2013 03:26 PM
To:  "ChenZhe"<cz at de3eb.cn>; 
Cc:  "sheepdog"<sheepdog at lists.wpkg.org>; 
Subject:  Re: [sheepdog] [PATCH v2] zk: actively remove itself after leave_cluster()is called

 
On 05/13/2013 03:07 PM, ChenZhe wrote:
> actively remove itself from zookeeper than waitting for timeout.

What problem does this patch solve? If a node is crashed, zookeeper need
wait for timeout to detect this. Even with this patch, the zookeeper
will also need to wait. And If we call leave_cluster() actively, the old
code can broadcast it immediately, no?

Thanks,
Yuan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.wpkg.org/pipermail/sheepdog/attachments/20130513/6d1c8980/attachment-0004.html>


More information about the sheepdog mailing list