Frankly, I'm surprised it runs at all. 650M dos in 2G of memory is very, very, very aggressive. To get it to run at all I'm guessing that you have turned off things like term vectors, position info and the like.
Anyway... Solr/Lucene cache various data as queries come in, so the scenario is this I think > Solr starts up > you fire a query or two at it and it tries to load caches then blows up with OOM errors. The fact that all your physical memory is used up is due to MMapDirectory and isn't a problem, see: http://blog.thetaphi.de/2012/07/use-lucenes-mmapdirectory-on-64bit.html But you're really putting more documents in these machines than I _ever_ expect to see in that little memory. I think you need to back up and do a sizing exercise to set expectations, see: https://lucidworks.com/blog/sizing-hardware-in-the-abstract-why-we-dont-have-a-definitive-answer/ Best, Erick On Wed, Jul 22, 2015 at 4:26 AM, Vishnu perumal <vis...@knackforge.com> wrote: > HI, > > I am running SOLR server ( solr version 4.10.3 & java version 1.7.0_67). > > Server Details : Amazon aws c3.xlarge server (7.5 GB RAM, 4vCPU). Linux > Ubuntu 14.04.1 LTS and installed Bitnami Apache Solr 4.10.3-0. Allocated > JVM memory minimum 1GB and maximun 2GB. The server has around 650 million > dcoucment indexed (Indexed field count is 37 & stored field count is 10). > > [image: Inline image 1] > > When the solr server starts, it works fine for the few queries (filter > query). But after running a few more select queries (SpatialSearch), it > takes longer to respond the queries and physical memory usage goes to full > (99.5%). > > [image: Inline image 2] > > and getting the java heap space error. > > > [image: Inline image 3] > > > Does it cache all the result? but I have disabled documents caching, field > caching. I am expecting that after query gets completed, the memory should > be released in a few seconds. But that is not happening. Do I miss anything > here? > > > I have atached the config file here for your reference. I must admit, I am > not that expert with Solr. > > > > > > Thankyou, > Vishnu Perumal, > vis...@knackforge.com > >