Török Edwin wrote: > On 2009-04-13 22:25, Jay Deiman wrote: >> Török Edvin wrote: >> >> [snip] >> >> >>>> Well, *I* couldn't find much of any use in the ktrace output. However, >>>> if someone else would like to take a look at the trace file, I've made >>>> it available at: >>>> >>>> http://janus.splitstreams.com/clamav-ktrace.out.bz2 >>>> >>>> It is about 91MB compressed. >>>> >>> Can you run this command, and make the output available at the same >>> location? (as described in that article linked above): >>> $ kdump -T -f ktrace.out | ./alloctrace.py >>> >>> P.S. the alloctrace.py script is available at >>> http://keramida.wordpress.com/2008/10/15/extracting-useful-info-from-freebsd-malloc-tracing/ >>> >> I ran alloctrace.py before posting the trace file. It gave me zilch for >> output. I did take a look at the dump and I found no calls within it to >> malloc(), which I thought was odd (and part of the reason that I posted >> the trace file). I have posted the full text output of the kdump as >> well to make things easy on those not using FreeBSD. That is at: >> >> http://janus.splitstreams.com/clamav-kdump.txt.bz2 > > I tried to reproduce a memory leak under FreeBSD, but failed so far, > please open a bugreport, > and tell us from where you obtained ClamAV (ports or built yourself), > and what configure flags you used.
Will do. I will try and include as much info as possible. > > Memory usage was between 111M - 125M. > Unfortunately I wasn't able to extract any useful data from the trace, > can you make sure you created the trace like this: > > *$ env MALLOC_OPTIONS='U' ktrace clamd* > $ *kdump -tu -f ktrace.out > kdump.out* I did run with the MALLOC_OPTIONS='U' when I did it. I'll run a new ktrace and kdump just to be sure and post download links to the trace and dump files as I did in the previous emails. > > > It would be useful if you could find the file(s) that cause the memory leak. That is a difficult task due to the sheer load. I am quite certain that there are a large number of files that are contributing to this as I currently have a nightly cron to restart all of the clamd processes on all 10 machines in the cluster. The memory usage is nearly 1GB on all machines at the restart time. Thanks for all the help so far -- Jay Deiman \033:wq! _______________________________________________ Help us build a comprehensive ClamAV guide: visit http://wiki.clamav.net http://www.clamav.net/support/ml
