On Apr 17, 2013, at 1:42 PM, Shawn Heisey <s...@elyograg.org> wrote:
> On 4/17/2013 10:29 AM, Umesh Prasad wrote: >> We use DIH and have turned off the Auto commit because we have to sometimes >> build index from Scratch (clean=true) and we not want to >> Our master server sees a lot of restarts, sometimes 2-3 times a day. It >> polls other Data Sources for updates which are quite a few. Master >> maintains a version of last committed version and can handle uncommitted >> changes. >> >> Given the frequent restarts, We can't really afford a huge start up at this >> point. >> In the worst case, does Solr allow for disabling transactional log ? > > Unless you are using SolrCloud, you can disable the updateLog. SolrCloud > requires it. > > There is one additional caveat - when you disable the updateLog, you have to > switch to MMapDirectoryFactory instead of NRTCachingDirectoryFactory. The > NRT directory implementation will cache a portion of a commit (including hard > commits) into RAM instead of onto disk. On the next commit, the previous one > is persisted completely to disk. Without a transaction log, you can lose > data. I don't think this is true? NRTCachingDirectoryFactory should not cache hard commits and should be as safe as MMapDirectoryFactory is - neither of which is as safe as using a tran log. - Mark > > My advice - keep the updateLog on, and use autoCommit with > openSearcher=false. It is the best way to avoid large transaction logs. It > sounds like you do not want the auto commits to affect query results, which > is a reasonable goal. You can have that even with autoCommit - just set > openSearcher to false. Here's an example, no need to stick with the numbers > that I have included: > > <updateHandler class="solr.DirectUpdateHandler2"> > <autoCommit> > <maxDocs>25000</maxDocs> > <maxTime>300000</maxTime> > <openSearcher>false</openSearcher> > </autoCommit> > <updateLog /> > </updateHandler> > > Thanks, > Shawn >
