[PVE-User] HDD errors in VMs

[Emmanuel Kasper]

On 01/07/2016 01:51 PM, Michael Pöllinger wrote:
> Hi Emmanuel.

Hi Michael

> Due to total crash of the VM there is no logging in that time period.
> So sadly we can´t provide any logs.
> 
> On sunday we kicked one node out of our cluster and resetup this node
> proxmox 4.1.
> Today the whole clusterserver crashed with no output on console. We just
> could reset it by ipmi.
> Now again up and running. Sadly our customers starts getting angry.
> (understandably)
> 
> Just sitting here and going through all logfiles for hours, but i can´t
> find anything.
> Only ONE strange entry in /var/log/pveproxy/access.log
> 115.231.222.14 - - [07/Jan/2016:04:05:40 +0100] "GET
> http://zc.qq.com/cgi-bin/common/attr?id=260714&r=0.4883031276012825
> HTTP/1.1" 501 -

This is unexpected indeed, but this should not crash the whole system.


> @4:24 during backup of one VM the whole node crashed.
> 
> I could understand the problem if:
> - it´s always to same times like backups or cronjobs
> - due to high load or errors on raid
> 
> but it´s always different times or on different load.
> sometimes VMs crashes by simply do nothing (idle)
> 
> Here a screenshot from the new proxmox 4.1 node
> http://imgur.com/GILhyw9
> 
> Some time ago there was a per incident ticket option. Have they gone
> completly?
> 
> Problems now:
> - Permanent sporadic crashes of debian 8.2 VMs
> - New one: whole crash of proxmox node.
> 

For this new one I very much suspect some hardware problem.

Yes, we stopped doing the per incident ticket option, but you alway can
get a standard suscription on your pve host, and downgrade that after
the one year period.

Remoting to your system would indeed allow to diagnose this much faster.


Emmanuel