zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeremy Stribling <st...@nicira.com>
Subject Re: znode metadata consistency
Date Tue, 12 Apr 2011 20:16:23 GMT
Yes, Zookeeper is running in our application (Cassandra is as well, in 
the same JVM).  We do this because we're building a product that will be 
deployed by customers with little to no distributed systems admin 
experience, and we only want them to have to manage a single process.  
We have not yet seen the need for knowing about the status of a 
particular application node independent of that node's Zookeeper status.

On 04/12/2011 01:12 PM, Ted Dunning wrote:
> Does this mean that you are running ZK in your application?
> My own feeling is that this is a bad practice.  It shouldn't cause 
> your problem, but it is hard to make guarantees about totally correct 
> operation in an arbitrarily shared JVM.
> In my own designs, I avoid collocating ZK with my application because 
> I like to interrogate ZK about application status and "down" is a 
> valid status which I would like to know about.  I can't see that in ZK 
> if it doesn't survive the demise of the application.
> On Tue, Apr 12, 2011 at 12:33 PM, Jeremy Stribling <strib@nicira.com 
> <mailto:strib@nicira.com>> wrote:
>     * We manage node membership dynamically, and our application
>     restarts the ZooKeeperServer classes whenever a new node wants to
>     join (without restarting the entire application process).  This is
>     why you'll see messages like the following in node1.0x4-0x5.log
>     before a new election begins:
>     2011-04-11 21:16:00,762 4804 [QuorumPeer:/
>     <>] INFO
>     org.apache.zookeeper.server.quorum.Learner  - shutdown called

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message