[SOLVED] Proxmox kills VMseven though a lot of memory is available

G

guacamole

New Member
Jun 4, 2025
3
0
1
I hope this hasn't answered before. I found a lot of threads about killing vms because of memory, but in all cases i found the assigned memory was simply greater than the total memory available. This is not the case here.

I have 3 vms on that machine

>machine 1: 14gb ram 12 cores
>machine 2: 2gb ram 2 cores
>machine 3: 4gb ram 2 cores

according to the resources tab the hardware has 27.19 GB RAM and 16 Cores. So lots of memory available

Now from time to time ( i talk about a few months with running without issues) proxmox shuts down my vms. The journal has the following entry

>Jun 04 06:11:32 pve kernel: Out of memory: Killed process 936779 (kvm) total-vm:9303024kB, anon-rss:4395844kB, file-rss:640kB, shmem->
>Jun 04 06:11:30 pve systemd[1]: 160.scope: A process of this unit has been killed by the OOM killer.
>Jun 04 06:11:30 pve systemd[1]: 160.scope: Failed with result 'oom-kill'.
>Jun 04 06:11:31 pve systemd[1]: 160.scope: Consumed 3d 6h 15min 39.054s CPU time.
>Jun 04 06:11:32 pve systemd[1]: 100.scope: A process of this unit has been killed by the OOM killer.
>Jun 04 06:11:32 pve systemd[1]: 100.scope: Failed with result 'oom-kill'.
>Jun 04 06:11:32 pve systemd[1]: 100.scope: Consumed 5d 6h 10min 33.114s CPU time.

any idea what could cause this? And especially what I can do to prevent this? I could reduce the memory but as there is already a lot of puffer I would assume this only lenghens the interval until this occurs again?

Thanks

(i am aware that the number of course is maybe to high but as the memory is the reason to shutdown i dont expect this to be the issue)
 
i added the complete dump of that timespan (you can ignore any "host not found" messages as the dns is on one of that killed machines)
 

Attachments

guacamole said:
So lots of memory available
Click to expand...
Actually no, the log shows that you have only about 200 MB available in your buddyinfo

Code: 
Jun 04 06:11:32 pve kernel: Node 0 DMA32: 1562*4kB (UM) 3212*8kB (UME) 2774*16kB (UME) 12*32kB (UM) 6*64kB (U) 5*128kB (UM) 4*256kB (UM) 3*512kB (UM) 2*1024kB (U) 11*2048kB (UM) 0*4096kB = 104872kB
Jun 04 06:11:32 pve kernel: Node 0 DMA32: 5436*4kB (UM) 3591*8kB (UME) 2847*16kB (UME) 0*32kB 0*64kB 4*128kB (UM) 4*256kB (UM) 3*512kB (M) 2*1024kB (UM) 2*2048kB (U) 0*4096kB = 105240kB

Just memory is no enough, you need congruent memory available, e.g. if the process requires e.g. a congruent 4 MB block, your system was unable to comply and the OOM ran.

Why don'tyou have swap configured?
 
Your last question seems to be part the answer. TBH I don't know why no swap was configured. Was a bit of a journey to create a swapfile in zfs but now it seems to work.
Code: 
Swapon -summary
and the web interface show the swapfile now. Added 32GB. Should be more than enough,

Thanks for your help!
 
Last edited:
  • Like
Reactions: guacamole and UdoB
You must log in or register to reply here.
Top Bottom