Re: Qemu also out of memory overnight
On 7/1/19 3:59 AM, wwp wrote:
Hello Robert,
On Sun, 30 Jun 2019 09:54:14 -0400 Robert Moskowitz <rgm(a)htt-consult.com> wrote:
> And the Fedora21 image is limited to 1GB:
>
> Jun 30 01:53:35 lx140e kernel: [ 13357] 107 13357 884212 215032 3043328
72575 0 qemu-system-x86
> Jun 30 01:53:35 lx140e kernel:
oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/machine.slice/machine-qemu\x2d3\x2dfedora21.scope,task=qemu-system-x86,pid=13357,uid=107
> Jun 30 01:53:35 lx140e kernel: Out of memory: Killed process 13357 (qemu-system-x86)
total-vm:3536848kB, anon-rss:860128kB, file-rss:0kB, shmem-rss:0kB
> Jun 30 01:53:35 lx140e kernel: oom_reaper: reaped process 13357 (qemu-system-x86),
now anon-rss:0kB, file-rss:12kB, shmem-rss:0kB
> Jun 30 01:53:36 lx140e journal[878]: internal error: End of file from qemu monitor
> Jun 30 01:53:36 lx140e systemd[1]: machine-qemu\x2d3\x2dfedora21.scope: Succeeded.
> Jun 30 01:53:36 lx140e systemd-machined[760]: Machine qemu-3-fedora21 terminated.
>
> A new kernel came out today, and I installed that and rebooted. So let's see
what happens tonight...
Looking at your other post, it seems that some process is requesting a
bunch of memory during the night, and the kernel kills the most memory
consumer ones to make room for the new one (and according to other
criteria). I had this exact same behavior when a rsync-based backup
was running overnight: it was listing files from an enormous disk, the
amount of memory for this was huge and firefox and other
big memory eaters being killed nearly every night.
No crashes this past night. Interesting you mention nighttime rsync.
My crontab has:
1 3 * * * rsync -tvz ftp.rfc-editor.org::rfcs/*.txt /home/common/ietf/rfcs
1 4 * * * rsync -tvz rsync.tools.ietf.org::tools.id/*.txt
/home/common/ietf/drafts
To get all the IETF rfcs and drafts. These two directories are HUGH! I
rsynced my old system to the new, but did not run this for a few days on
the new system. Plus this is the rush time before an IETF meeting (IETF
105 July 21st), and everyone is crunching on their drafts and the RFC
editor is pushing to get the approved RFCs published.
So it COULD be that. I will have to monitor this and see. And how to
get the system to just put on or how to limit memory used by rsync.