We recently upgraded our cloud from 4.8 to 4.10.3, the only config we updated was the luceneMatchVersion. Response times were very stable prior to the upgrade, but are quite erratic since the upgrade, and rising. I still have to check all the resolved issues but something went very wrong between 4.8 and 4.10.3.
M. -----Original message----- > From:Toke Eskildsen <t...@statsbiblioteket.dk> > Sent: Wednesday 4th February 2015 20:58 > To: solr-user@lucene.apache.org > Subject: RE: low qps with high load averages on solrcloud > > Suchi Amalapurapu [su...@bloomreach.com] wrote: > > Noticed that a solrcloud cluster doesn't scale linearly with # of nodes > > unlike the unsharded solr cluster. We are seeing a 10 fold drop in QPS in > > multi sharded mode. > > As I understand it, you changed from single to multi shard. > > Guessing wildly: You have one or more facets with a non-trivial (10K or more) > number of unique String values and you have a fairly high facet.limit (50+). > If so, what you see might be the penalty for the two-phase faceting with > SolrCloud, where the second fine-counting phase can be markedly slower than > the first. There are ways to help with that, but let's hear if my guess is > correct first. > > - Toke Eskildsen >
