From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040115 Description of problem: I have several compute nodes with 6GB of RAM and 2GB of swap, running primarily ls-dyna, a commercial FEM code compiled for AMD64. On one of these, a simulation was actively running with a VSS of 2274180 and a RSZ of 652240 (almost a quarter of it is swapped out!). Another user tried to fire up a simulation that wanted 3371384 of VSS. The OOM killer swiftly terminated it. Should that happen? Any way to make it not happen? Version-Release number of selected component (if applicable): kernel-2.4.21-9.0.3.ELsmp How reproducible: Didn't try Additional info: I've set /proc/sys/vm/inactive_clean_percent to 30 in response to bug 115438 and /proc/sys/vm/pagecache to "1 15 15" in response to bug 118152.
Created attachment 102419 [details] Memory graph created by Ganglia
Created attachment 102420 [details] CPU graph created by Ganglia.
Upon further investigation, there may have been enough memory pressure on the system to trigger OOM after all. I'm going to close this, since I can't be completely clear on what happened. Sorry for the noise.