As far as I can tell (and please someone correct me if I'm wrong), currently it's best to restart your Solr servers (probably on a rolling basis) pointing to the new ZK machine and not pointing at the old one. I believe there's some work afoot both with ZK and Solr to be more robust in this situation, but the Solr work depends on the ZK work I think.
Although I wouldn't expect the ZK machines to be that fragile, or are you thinking about things like AWS or Azure etc? Erick On Mon, Oct 15, 2012 at 1:39 PM, John Brinnand <jbrinn...@gmail.com> wrote: > Hi Folks, > > I have been looking at solrcloud to solve some of our problems with solr in > a distributed environment. As you know, in such an environment, every > instance of solr or zookeeper can come into existence and go out of > existence - at any time. So what happens if instances of ZK disappear and > re-appear with different hostnames and DNS entries? How would solr know > about these instances and how would it re-sync with these instances? > > In essence my question is: what if the hostname and port of the ZK instance > no longer exists - how will solrcloud discover the new instance(s)? > > Thanks, > > John > > > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/solrcloud-what-if-ZK-instances-are-evanescent-tp4013740.html > Sent from the Solr - User mailing list archive at Nabble.com.
