On Thu, 2007-28-06 at 11:05 -0700, Mike Klaas wrote: > On 28-Jun-07, at 11:02 AM, Yonik Seeley wrote: > > > On 6/28/07, Michael Thessel <[EMAIL PROTECTED]> wrote: > >> I got a solr server running to index the posts of a forum. The > >> server is > >> running in a tomcat installation on a 4-processor Opteron server. > >> One of > >> the threads is eating up 100% of one processor all the time. > > > >> I do auto-commits every 10s > > > > It's probably autowarming (pre-populating new caches based on the > > contents of the current caches). If your index is a reasonable size, > > and you aren't doing anything like faceting, then try adjusting down > > the autowarm counts (set to 0 for the queryResultCache). The > > filterCache is trickier... you can try setting that autowarm count to > > 0 and see if it negatively affects your queries. > > Huh? That would be surprising since the OP mentioned it happens all > night when no new docs are added. > > Michael, have you configured autoCommit, or are you programmatically > sending <commit/> every 10s?
I configured autoCommit which works pretty well: <updateHandler class="solr.DirectUpdateHandler2"> <autoCommit> <maxDocs>1000</maxDocs> <maxTime>10000</maxTime> </autoCommit> </updateHandler>