Hide Forgot
Description of problem: gnome session gets killed when system runs out of memory. No warning message is raised just all programs are killed and session is restarted. Version-Release number of selected component (if applicable): gnome-session-2.28.0-17.el6 How reproducible: Few minutes after oom (free memory about 5%) Steps to Reproduce: 1. run Firefox 5.0 with a few tabs (more is better) OR run another memory eater 2. gnome-session will be restarted without any warning Actual results: system will freeze for few minutes then running programs will be killed, session will restart. Log: [alich@angband ~]$ dmesg | grep gnome-session Out of memory: kill process 2754 (gnome-session) score 353386 or a child Out of memory: kill process 2754 (gnome-session) score 341243 or a child Out of memory: kill process 2754 (gnome-session) score 318305 or a child Out of memory: kill process 2754 (gnome-session) score 293765 or a child Out of memory: kill process 2754 (gnome-session) score 279522 or a child Out of memory: kill process 2754 (gnome-session) score 224367 or a child Out of memory: kill process 2754 (gnome-session) score 215884 or a child Out of memory: kill process 2754 (gnome-session) score 208070 or a child Out of memory: kill process 2754 (gnome-session) score 201137 or a child Out of memory: kill process 2754 (gnome-session) score 193052 or a child Out of memory: kill process 2754 (gnome-session) score 192182 or a child Out of memory: kill process 2754 (gnome-session) score 176468 or a child Out of memory: kill process 2754 (gnome-session) score 166597 or a child Out of memory: kill process 2754 (gnome-session) score 135099 or a child gnome-session invoked oom-killer: gfp_mask=0x201da, order=0, oom_adj=0 gnome-session cpuset=/ mems_allowed=0 Pid: 2754, comm: gnome-session Tainted: G ---------------- T 2.6.32-169.el6.x86_64 #1 Out of memory: kill process 2754 (gnome-session) score 80745 or a child Out of memory: kill process 2754 (gnome-session) score 73733 or a child Out of memory: kill process 2754 (gnome-session) score 61517 or a child Out of memory: kill process 2754 (gnome-session) score 29461 or a child Out of memory: kill process 15963 (gnome-session) score 456454 or a child Out of memory: kill process 15963 (gnome-session) score 438191 or a child Out of memory: kill process 15963 (gnome-session) score 423777 or a child Out of memory: kill process 15963 (gnome-session) score 377830 or a child Out of memory: kill process 15963 (gnome-session) score 347221 or a child Out of memory: kill process 15963 (gnome-session) score 298338 or a child Out of memory: kill process 15963 (gnome-session) score 286850 or a child Out of memory: kill process 15963 (gnome-session) score 276432 or a child Out of memory: kill process 15963 (gnome-session) score 267189 or a child Out of memory: kill process 15963 (gnome-session) score 256209 or a child Out of memory: kill process 15963 (gnome-session) score 255049 or a child Out of memory: kill process 15963 (gnome-session) score 233821 or a child Out of memory: kill process 15963 (gnome-session) score 220965 or a child Out of memory: kill process 15963 (gnome-session) score 150343 or a child Out of memory: kill process 15963 (gnome-session) score 111259 or a child Out of memory: kill process 15963 (gnome-session) score 99742 or a child Expected results: Warning message, no program is killed Additional info: I have 2GB of RAM and about 150MB free when it starts the carnage process.
we could fool around with oom_adj or so but that won't solve the problem generally, and so I don't think it's really worth doing. I think the answer here is: 1) if you're going to run a system with only 2gb of ram, enable swap 2) if you don't want the default kernel behavior of "kill an arbitrary process" then manually set overcommit as appropriate What are your thoughts?
Hi Ray, My choise is 1. But I'm so surprise because I thought I had a swap on! :O Actually, I don't know how is possible that I haven't it because I would never use the desktop without swap. Thanks for advice I'll fix it and hope it solve the problem.
Well it seems that SWAP was turned on but I have to say it's something strange in there. I use LVM and have 3 root partitions...when I want to test other release "life" I use the old one. At last there is swap on but it seems it's something strange in there - I need some time to investigate it.
Since RHEL 6.2 External Beta has begun, and this bug remains unresolved, it has been rejected as it is not proposed as exception or blocker. Red Hat invites you to ask your support representative to propose this request, if appropriate and relevant, in the next release of Red Hat Enterprise Linux.
Aleš, what's the bug number for the firefox issue ?