Latest pve kernel has a fix for a serious oom killer bug. I would try upgrading your kernel before anything else.
On February 4, 2017 3:16:05 PM GMT+01:00, Michele Bonera <[email protected]> wrote: >On 04/02/2017 12:35, Alwin Antreich wrote: > >> Hi Michele, >> >> On 02/04/2017 10:44 AM, Michele Bonera wrote: >>> Hi. >>> >>> I have an issue with OOM Killer (Proxmox 4.4-5 - Kernel >4.4.35-1-pve) on >>> my infrastructure: even if there is a lot of free memory (15GB used >over >>> 32GB available), OOM Killer is still killing my VM processes. >> >> Are you over committing your machines? > >Not at all: the total amount of RAM assigned to the VMs running on this > >node is 12GB (and the host has 32GB). > >> Another consideration could be NUMA, if one node runs out of memory >the >> kernel might also kill processes. > >This is plausible. I'll do some checks... Thanks for the hint. > >>> This morning I had 3 over 6 VM powered down because OOM Killer >decided >>> to kill the related processes even if there were a lot of free >memory. >>> >>> total used free shared buffers cached >>> Mem: 32725572 32286280 439292 196428 3772084 >18648708 >>> -/+: 9865488 22860084 >>> Swap: 1499124 33736 1465388 >> >> It could also be possible that your swap space can't be utilized >> properly due to your workload. > >I exclude this: the cpu utilization rarely exceed 20% > >Bye >-- >Michele Bonera >www.zipoware.com >_______________________________________________ >pve-user mailing list >[email protected] >http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. ---- This mail was virus scanned and spam checked before delivery. This mail is also DKIM signed. See header dkim-signature. _______________________________________________ pve-user mailing list [email protected] http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
