Ok, so we have 2 different scenarios here, running a query through admin UI and I'm assuming there you are quoting qTimes from the )UI, direct to the relevant core, and running through CloudSolrServer (what you called "running a query through ZK").
Questions: - What timings are you using when you run through CloudSolrServer? Are you measuring elapsed time yourself, or getting qTimes by parsing Qtime from the response? If you are measuring elapsed time, then you are comparing apples with pears. Check the qTimes in the responses, those should be equivalent to what the UI shows. - What parameters do you use to CloudSolrServer? Do you enable parallel updates, updates to leader, etc? I know these are queries but would be good to know what options you have set (if any). - What else is going on in your application when you talk to CloudSolrServer, could it be a GC pause in your app? On 16 June 2014 17:23, drmangrum <drmang...@yahoo.com> wrote: > I'm using SolrJ's CloudSolrServer > < > http://lucene.apache.org/solr/4_8_0/solr-solrj/org/apache/solr/client/solrj/impl/CloudSolrServer.html > > > that accepts the address of your zookeeper instances and does the grunt > work > for you. > > > I guess the problem could be with that class. > > > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/Inconsistent-query-times-tp4138956p4142091.html > Sent from the Solr - User mailing list archive at Nabble.com. >
