When my system ran out of memory, the oom-killer got active and did the following:

Jan 22 10:26:18 linux kernel: Out of Memory: Killed process 8280 (klauncher).
Jan 22 10:26:46 linux kernel: Out of Memory: Killed process 8305 (kwin).
Jan 22 11:32:15 linux kernel: Out of Memory: Killed process 8320 (xterm).
Jan 22 11:58:25 linux kernel: Out of Memory: Killed process 8842 (firefox-bin).
Jan 22 11:58:31 linux kernel: Out of Memory: Killed process 8321 (xterm).

Note the total time it took to recover, and especially note the innocent processes killed before the very likely main culprit (firefox) was hit. There was also a konqueror process with >600 MB (thats too large, even for konqueror) on the system which I killed manually to prevent further oom situations. 

I tried to connect to the machine from around 11:50, and wasn't successful until 12:05, I can't comment on the responsiveness before that time. 

I'm using 10.0 (kernel 2.6.13-15.7-smp) on a machine with 1GB RAM and 1GB swap. The system was used as a desktop machine, with a single locked KDE session and me trying to connect via ssh.


This problem started a discussion on [kernel], during which Kurt submitted a patch. As the discussion is ongoing, I still submit this bug for reference and closing once the discussion about oom mechanics is finished.