We are seeing the following error frequently in our catalina.out.
Occasionally, perhaps once a day on average, we see these errors thrown in large numbers (several hundred a minute). Our application then slows to a crawl and only a restart of tomcat will fix the issue. It then continues to run ok in the short term.
Has anyone else had this problem, or knows what might be causing it?
Env --- Tomcat 6 (6.0.16), Java 6 (1.6.0_05). APR 1.2 mod_jk (mod_jk/1.2.6) Linux [ May 02, 2008: Message edited by: Charles Craft ]
Joined: May 02, 2008
Just a follow up incase anyone else comes across this post.
Although I cannot be sure as we haven't seen this error for a few days now, I think it was down to low memory/GC overhead which we have hopefully fixed, I will also put the memory fix information below as it took some time to find the culprit and may help if people are googling.
When profiling our application using jmap and the -histo option, I was seeing huge int (shown as [I by jmap), in the order of 700-800 MB, running with about 1.5GB of allocated memory.
Using jhat this was traced back to SuballocatedIntVector, a xalan class. There appeared to be a huge memory leak here. We use XSLT extensively. I upgraded to xalan 2.7.1 (previously using xalan 2.5.1) and this has completely solved this particular issue. I'm now seeing int's in the order of 10MB.