Another thing is to put in some autowarming, both on the caches and firstSearcher and newSearcher. These will pre-fill the caches before having new searchers handle queries.
Don't go overboard here, for things like filterCache try, say, autowarm counts of 16 and work your way up. firstSearcher is used when you restart Solr and don't have any caches to run autowarm on. newSearcher is run when you commit, and presumably your caches have entries to autowarm from. Try to put things that sort on fields you know you're going to sort on, some of the facets you expect etc. Best, Erick On Fri, Nov 7, 2014 at 8:46 AM, Toke Eskildsen <t...@statsbiblioteket.dk> wrote: > mizayah [miza...@gmail.com] wrote: >> What i see is that after restart jvm usage is realy low and raise slowly >> while system cpu ussage is high. >> My select queries are realy slow during that time. > > The first searches tend to be slow while Solr fills internal caches and the > OS file cache is warmed. > >> After few days when jvm grab some MORE memory system drops down. >> java settings >> -Xmx6144m -XX:+UseConcMarkSweepGC -XX:+PrintGC -XX:+PrintGCDetails >> I have 8GB ram > > I am guessing you are using spinning drives? You have 8GM RAM in your > machine. Let's say your JVM uses 3GB at the beginning and the rest of your > system 1GB. That leaves 4GB for disk cache. When the JVM grabs more memory > and increases the heap, let's say to 5GB, that leaves only 2GB for disk > cache. As Solr needs fast I/O, either in the form of fast storage or enough > free memory for OS file cache, your performance goes down. > > Possible solutions: > > 1) Set Xmx as low as practically possible, leaving as much memory as possible > for file cache > 2) Put more RAM in the machine > 3) Switch to SSD > > How large is your index? > > - Toke Eskildsen
