Re: SolrCloud - load balancing
Mark Miller <markrmiller <at> gmail.com>
2012-08-04 01:50:51 GMT
Okay - it does basically cycle through the servers. I forgot, it shuffles the server list every time.
So it's okay for searching - but we want to make it smarter for indexing.
First perhaps by favoring leaders, and then hashing itself and sending directly to the right leader.
Sent from my iPhone
On Aug 3, 2012, at 8:35 PM, Mark Miller <markrmiller <at> gmail.com> wrote:
> Right - though I'm not sure it load balances offhand - I think it might just do the fail over part rather than
the round robin . I was looking at it the other day and need to investigate.
> I also think it should be weighting towards sending to leaders - eg ideally it should load balance across
the leaders for indexing and everyone for search, round robin.
> On Fri, Aug 3, 2012 at 7:49 PM, Jack Krupansky <jack <at> basetechnology.com> wrote:
> You can use CloudSolrServer - "SolrJ client class to communicate with SolrCloud. Instances of this class
communicate with Zookeeper to discover Solr endpoints for SolrCloud collections, and then use the
LBHttpSolrServer to issue requests."
> -- Jack Krupansky
> -----Original Message----- From: sausarkar
> Sent: Friday, August 03, 2012 7:29 PM
> To: solr-user <at> lucene.apache.org
> Subject: SolrCloud - load balancing