[PVE-User] Kernel Memory Leak on PVE6?
Stoiko Ivanov
s.ivanov at proxmox.com
Wed Oct 2 16:20:21 CEST 2019
hi,
On Wed, 2 Oct 2019 15:54:05 +0200
Chris Hofstaedtler | Deduktiva <chris.hofstaedtler at deduktiva.com> wrote:
> Replying to myself once more, mostly for the benefit of people
> following along at home...
>
> * Chris Hofstaedtler | Deduktiva <chris.hofstaedtler at deduktiva.com> [191002 00:07]:
> > > root at vn03:~# uname -a
> > > Linux vn03 5.0.21-1-pve #1 SMP PVE 5.0.21-1 (Tue, 20 Aug 2019 17:16:32 +0200) x86_64 GNU/Linux
> >
> > I've upgraded both machines yesterday to:
> > Linux vn03 5.0.21-2-pve #1 SMP PVE 5.0.21-6 (Fri, 27 Sep 2019 17:17:02 +0200) x86_64 GNU/Linux
> >
> > And they seem to be doing fine for now. Also, the slab size is a lot
> > smaller just after boot, compared to previous reboots.
>
> So, that helped, but it's not the entire story.
>
> For anyone affected:
> - Do you have ipmitool, libfreeipmi17, freeipmi-common installed?
> - Are you running check_mk agent or another tool polling ipmitool
> data?
nice catch!! It'd seem one other cases (e.g. [0]) also have check_mk_agent
running (otoh one reproducer has this on a KVM-Guest).
In any case I'll share that as potential source!
Thanks!
[0] https://forum.proxmox.com/threads/pve6-slab-cache-grows-until-vms-start-to-crash.58307/
> - Can you try uninstalling ipmitool libfreeipmi17 freeipmi-common
> to see if the problem goes away?
>
> Plus, if you happen to have check_mk and can't see details in your
> memory allocation graphs, you can add an extra "slab only" graph
> by editing:
> /opt/omd/versions/1.5.0p22.cre/share/check_mk/pnp-templates/check_mk-mem.linux.php
> and adding:
> $opt[] = $defopt . "--title \"Slab only\"";
> $def[] = ""
> . mem_area("slab", "af91eb", "Slab (Various smaller caches)", FALSE)
> ;
> starting at line 95.
>
> Chris
>
More information about the pve-user
mailing list