On Mon, Jul 13, 2009 at 7:26 PM, Jason Rutherglen<jason.rutherg...@gmail.com> wrote: > Has anyone looked at implementing automatic failover in SOLR using a naming > service (like Zookeeper)?
Yep, zookeeper integration would be great! - optionally provide a single point to configure a cluster (i.e. load solrconfig.xml, schema.xml, etc from there) - track what machines are in the cluster, and what shards they contain We could get rid of the currently required load balancer for fault tolerant setups and have a SolrJ client that looks at zookeeper for what servers it should query. Actually, it would be nice to also provide that functionality in Solr itself, so clients could optionally be very simple... if the collection is distributed, Solr could turn the request into a distributed one and query the correct shards. I think we should start this after 1.4 gets out the door. -Yonik http://www.lucidimagination.com
