It's a known issue:
https://issues.apache.org/jira/browse/ZOOKEEPER-1197 that has no real
solution (although feel free to try one of the patches on 1197) due to
a limitation of TCP and a flawed implementation on our side.
It's being addressed here: https://issues.apache.org/jira/browse/ZOOKEEPER-1346
On Wed, Dec 19, 2012 at 7:30 AM, Kristopher Kane <[EMAIL PROTECTED]> wrote:
> I'm running Zookeeper 3.3.3 on CHD3u1 (RHEL 5 - JRE 1.6.0_33) in three
> ensembles for three different application sets.
> One ensemble is for SolrCloud. Watching zktop, we see it flail constantly
> with unknown states of the three zookeepers.
> This is similar to sending 'stat' and information apparently being cut off
> upon return to the calling client.
> What does it say about the health of ZK if the full response is not
> returned on the request of 'stat' ?
> -Kris Kane