One additional thought here: from a paranoid risk-management perspective it's not a good idea to have two critical services dependent upon a single point of failure if the hardware fails. Obviously risk-management is suited to taste, so you may feel the cost/benefit does not merit the separation. But it's good to make that decision consciously…you'd hate to have to justify a failure here after-the-fact as something overlooked :)
On Aug 30, 2013, at 9:40 AM, Shawn Heisey <s...@elyograg.org> wrote: > On 8/30/2013 9:43 AM, Jared Griffith wrote: >> One last thing. Is there any real benefit in running SolrCloud and >> Zookeeper separate? I am seeing some funkiness with the separation of the >> two, funkiness I wasn't seeing when running SolrCloud + Zookeeper together >> as outlined in the Wiki. > > For a robust install, you want zookeeper to be a separate process. It can > run on the same server as Solr, but the embedded zookeeper (-DzkRun) should > not be used except for dev and proof of concept work. > > The reason is simple. Zookeeper is the central coordinator for SolrCloud. > In order for it to remain stable, it should not be restarted without good > reason. If you are running zookeeper as part of Solr, then you will be > affecting zookeeper operation anytime you restart that instance of Solr. > > Making changes to your Solr setup often requires that you restart Solr. This > includes upgrading Solr and changing some aspects of its configuration. Some > configuration aspects can be changed with just a collection reload, but > others require a full application restart. > > Thanks, > Shawn >
