Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Accumulo, mail # user - Tserver kills themselves from lost Zookeeper locks


Copy link to this message
-
Re: Tserver kills themselves from lost Zookeeper locks
Josh Elser 2013-11-05, 00:42
On 11/4/13, 6:52 PM, Eric Newton wrote:
> Hmm.  That's sounds like a very old version of zookeeper... the earliest
> version we recommend is 3.3.6.
> I have never even heard of zookeeper version 2.1.3.2.  Is it specific to
> the MapR environment?

That looks like a version of MapR, not ZooKeeper. MapR's release notes
don't list the version for us either:
http://doc.mapr.com/display/RelNotes/Version+2.1.3.2+Release+Notes

>
> Virtual environments can be tricky.  There's a second level of possible
> swap that can take place, so it's possible you are getting something
> pushed out of memory: either the tserver or a virtual OS running the
> tserver.
>
> You may want to look into increasing your zookeeper timeouts in this
> environment.

 From the Accumulo side: `instance.zookeeper.timeout` in
accumulo-site.xml. That variable defaults to 30 seconds and controls the
timeout length for ZooKeeper operations.

>
> -Eric
>
>
>
>
>
> On Mon, Nov 4, 2013 at 4:08 PM, buttercream
> <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>
> wrote:
>
>     I'm running a 3 node development environment cluster, 1 master and 2
>     tablet
>     servers. I'm running Accumulo 1.5 with MapR M3. Zookeeper version
>     2.1.3.2. I
>     am in a virtual environment running RHEL 6.
>
>     It usually takes about 5 days after starting the cluster for the
>     tservers to
>     start dying.
>
>
>
>     --
>     View this message in context:
>     http://apache-accumulo.1065345.n5.nabble.com/Tserver-kills-themselves-from-lost-Zookeeper-locks-tp6125p6131.html
>     Sent from the Users mailing list archive at Nabble.com.
>
>