On Thu, Jun 18, 2009 at 8:30 PM, Peter Wolanin<peter.wola...@acquia.com> wrote: > So for now would it make sense to spread out the autocommit times for > the different cores?
Sure. You might also consider using commitWithin (solr 1.4) when updating the index - then you could either send the updates at slightly different times, or add a random amount of time to the commitWithin for each update. -Yonik http://www.lucidimagination.com > Thanks. > > -Peter > > On Thu, Jun 18, 2009 at 7:07 PM, Yonik Seeley<yo...@lucidimagination.com> > wrote: >> On Thu, Jun 18, 2009 at 4:27 PM, Peter Wolanin<peter.wola...@acquia.com> >> wrote: >>> I think I understand >>> that all the pending changes are on disk already, so the "commit" that >>> happens when the time is up is really just opening new searchers that >>> include the added documents. >> >> Only some of the pending changes may be on disk - a solr level commit >> involves closing the IndexWriter which flushes everything to disk, and >> then a new IndexReader is opened to read those changes. >> >> This will be improved in future versions such that an IndexReader can >> be opened *before* all of the changes have been flushed to disk (work >> on near-real-time indexing/searching in Lucene is progressing). >> >> -Yonik >> http://www.lucidimagination.com >> > > > > -- > Peter M. Wolanin, Ph.D. > Momentum Specialist, Acquia. Inc. > peter.wola...@acquia.com >
