[ 
https://issues.apache.org/jira/browse/MESOS-3391?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14935757#comment-14935757
 ] 

Vinod Kone commented on MESOS-3391:
-----------------------------------

[~jamespeach] To be clear this patch makes sure if a ZK server sends out of 
order pings, then it doesn't break the ZK C client used by mesos (scheduler 
driver/master/agent). It should not affect Mesos behavior if you are using 
unbundled ZK libraries and your ZK server is well behaved (no out of order 
pings). In other words, this change is backwards compatible.

[~neilc] We could wait for the official ZK release (not sure about the ETA) 
with the patch and just patch Twitter's Mesos internally (because Twitter's ZK 
servers send out of order pings for some optimization wins IIUC). But we were 
trying to upstream this patch in the hopes that it is a generally useful patch 
for the project/community to make mesos work with a wider variety of ZK servers.

[~rgs] [~cchen] Are you aware of any other ZK server implementations that do 
out of order pings? Are there any other compatibility concerns with this patch?

> Include patch for ZOOKEEPER-2253 for built-in Zookeeper 3.4.5 distribution
> --------------------------------------------------------------------------
>
>                 Key: MESOS-3391
>                 URL: https://issues.apache.org/jira/browse/MESOS-3391
>             Project: Mesos
>          Issue Type: Bug
>          Components: general
>         Environment: Linux, OS X
>            Reporter: Chris Chen
>            Assignee: Chris Chen
>
> The Zookeeper C client does makes certain assertions about the ordering of 
> ping packets that the Java client does not. An alternate implementation of 
> the Zookeeper server would then break the C client while working correctly 
> with the Java client.
> A patch has been submitted to the Zookeeper project under ZOOKEEPER-2253. 
> This adds that patch to mesos 3rdparty.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to