Proxmox 4.4.5 kernel: Out of memory: Kill process 8543 (kvm) score or sacrifice child

[whitewater]

It must not be for nothing that they return back
For me, no problem with the kernel test 4.4.35-2-pve and the 4.4.21-71 since more than one day.
So it was that.

 

[2beers]

Test kernel 4.4.35-2-pve fixes the problem for me, too.

 

[GPLExpert]

Hi,

Where we can get the kernel 4.4.35-2-pve please ?

 

[reetp]

Backup with new settings as above with same kernel worked for me last night

 

[2beers]

Hi,

Where we can get the kernel 4.4.35-2-pve please ?

1. Download
2. Install

 

[Jorge Teixeira]

So far so good with the test kernel.

 

[Marcus Dewald]

Thanks for the Feedback Jorge! I've just installed the test kernel too. Crossing fingers that Proxmox now stops to (blood)sacrifice my Windows VM.

 

[GPLExpert]

Thanks for the feedback and for answers

 

[fireon]

Upgraded 2 Cluster last night. Same problem, 4 VMs crashed. I test the new kernel too. Must i set the settings in systctl too?

 

[ozgurerdogan]

I have edited sysctlt and it has been fine for 5-6 days. But some other new nodes started same issue. I will probably need to edit all of nodes. But note sure if it is good fix.

 

[Marcus Dewald]

Thanks for the Feedback Jorge! I've just installed the test kernel too. Crossing fingers that Proxmox now stops to (blood)sacrifice my Windows VM.

In reply of myself and all whom may be concerned. The Test Kernel fixed my issue, VM is running since then.

 

[fireon]

In reply of myself and all whom may be concerned. The Test Kernel fixed my issue, VM is running since then.

Looks like the same, i've tested backups every 2 hours, and it seems to work.

 

[whitewater]

For me, only update kernel is enough. Not need to change something else.

 

[fabian]

kernel with the two offending commits reverted has been uploaded to pvetest and pve-no-subscription:
http://download.proxmox.com/debian/...4/pve-kernel-4.4.35-2-pve_4.4.35-78.changelog

it is essentially identical to the test build from earlier in this thread, minus the version numbers.

 

[whitewater]

ok, thank you Fabian.

 

[maxprox]

Hi their,
if you will work with the min_free_kbytes setting
have a look at the redhat doku and their Warning:
https://access.redhat.com/documenta...rformance_Tuning_Guide/s-memory-tunables.html
If I determine the value, as described here
https://my.vertica.com/docs/7.1.x/H...ionGuide/BeforeYouInstall/min_free_kbytes.htm
I get => 32463
(where I do not want to set this value at the time)

 

[c0urier]

I had the same issue with kernel : 4.4.35-1-pve and it was resolved by upgrading to : 4.4.35-2-pve

Tried with a quite large backup afterwards:
INFO: transferred 3543348 MB in 17702 seconds (200 MB/s)
INFO: archive file size: 2836.98GB
INFO: Finished Backup of VM 100 (04:55:06)
INFO: Backup job finished successfully
TASK OK

 

[Dan Nicolae]

Problem on our side was with the CEPH nodes. From time to time, OSD daemons where killed and CEPH marked them down. In one morning, half of our OSD was down, cluster was rebuilding, most of the VM partitions where corupted, some of them impossible to recover, data loss, absolutely horror. After the kernel update everything looks ok. Update took place on 3 january, from that time (7 days), a lot of pain... What a mess.

 

[fireon]

Thanks for the fast Kernelupdate in Enterpriserepo!

 

[snpz]

Hi there!

We have the same issue on our server cluster.
Fabian - is this bug fixed in pve-kernel-4.4.35-2-pve that is available using dist-upgrade or it is better to install the version you built?
Thanks in advance!