I would also get some metrics when SOLR is doing nothing, the JVM does do work in the background and looking at the memory graph in VisualVM will show a nice sawtooth.
François On Aug 14, 2014, at 1:16 PM, Erick Erickson <erickerick...@gmail.com> wrote: > bq: I just don’t know why Solr is suddenly going nuts. > > Hmmm, as Shawn says, hard to say at this remove. But > I've personally doubled the memory requirements for Solr > on the _same_ index by altering the query to a pathological > one. Something like > q=*:*&facet.field=whatever > where the field "whatever" contains a billion unique strings is > an example of a pathological query. > > So you may have to do the ugly work of correlating memory spikes > with the queries just prior to the spike. Which you should be able > to do from the Solr logs. > > Sorry I can't be more help... > Erick > > On Thu, Aug 14, 2014 at 9:45 AM, Shawn Heisey <s...@elyograg.org> wrote: >> On 8/14/2014 10:06 AM, Scott Rankin wrote: >>> My question was actually more about what in Solr might cause the >>> server to suddenly go from a very consistent heap size of 300-400 MB >>> to over 2 GB in a matter of minutes with no changes in traffic. I get >>> why the VM is crashing, I just don’t know why Solr is suddenly going nuts. >> >> That's nearly impossible to answer. Chances are that something has >> changed about the requests that Solr is receiving and now it's required >> to do something that it wasn't before, something that uses a lot of heap >> memory. >> >> The other likely possibilities are: >> >> * There's a bug in your solr version or in some software component that >> you are using with Solr. That can include the Java virtual machine, the >> servlet container, and/or any third-party Solr components. >> >> * You were running on the hairy edge of heap usage already, and >> something (a traffic increase, a slight change to your requests) pushed >> you over the edge into OutOfMemory. >> >> Thanks, >> Shawn >>
