On Monday 10 May 2004 17:26, Remy Maucherat wrote: > Stefan Proels wrote: > > Dear all, > > > > after updating our server to Tomcat 5.0.24 last night I got 2 crashes > > with the last words > > > > Exception in thread "CompilerThread1" java.lang.OutOfMemoryError: > > requested 32760 bytes for ChunkPool::allocate. Out of swap space? > > > > from Tomcat. Now the interesting is that according to our GC logs > > Java did not really run out of memory. Also I checked the system's > > state some minutes before the last crash and I remember the java > > process having allocated about 270 MB and as I set -Xmx1024m this > > also makes be believe that it did not really run of the memory. The > > pointer to the swap space also doesn't seem to lead into the right > > direction since the machine has 4 GB physical memory and another 4 GB > > swap space and nearly no swap space is actually in use. > > > > So my question is, is there any new configuration parameter I my have > > overlooked since 5.0.19 or has the default value of some related > > parameter changed which could cause this error? > > > > What makes the problem even more mysterious is the fact that both > > exceptions occurred in "CompilerThread1" but since all our JSPs have > > been compiled already and none of them has changed, I wonder why this > > thread had anything to do at all. > > You can (= should) disable the compiler thread IMO. > If you're precompiling every JSPs to servlets (with mappings in > web.xml), you should leave Jasper in development mode, and it won't > spawn the thread.
Sorry, I wasn't precise enough here. I'm not statically precompiling JSPs that way. Sometimes we do need the feature to update a JSPs while the application is running. I'm running a simple script which invokes every JSP with jsp_precompile=true instead and since this script already ran I wonder what "CompilerThread1" still has to do. I've had several more crashes, all with an OutOfMemoryError in "CompilerThread1" without the Java process actually running out of heap. I finally went back to 5.0.19 for now. I still hope there is a possibility to fix this problem somehow, because 5.0.24 contains several bug fixes I was waiting for and thus I'd like to update. Best regards, Stefan --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
