[Pkg-ganeti-devel] Bug#875435: ganeti: random freeze on some kvm guests in a jessie ganeti cluster

[Apollon Oikonomopoulos]

Control: tags -1 + moreinfo

Hi,

On 13:33 Mon 11 Sep     , Thomas Pierson wrote:
> Package: ganeti
> Version: 2.15.2-1~bpo8+1
> Severity: normal
> 
> Dear Ganeti maintainers,
> 
> I get random freeze on kvm guests in a Jessie Ganeti cluster.
> 
> Last time it happened 2 times on the same guest at one week of interval.
> Each time the guest only give ping response but nothing more and it get
> back to work only after a `gnt-instance reboot`.
> 
> I did not find trace or error in log files in the guest or in the hosts.
> 
> I use ganeti and drbd packages from jessie-backports and kvm and kernel
> packages from jessie archive.
> The guest is a Debian Wheezy system but it got trouble on Debian Jessie
> guest too.
> One month ago I already post a bug report against the linux package
> about disk lock on some Jessie guests in the same cluster:
> - https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=870471
> In that cases I got something on the guests kernel logs so maybe it's a
> different issue, maybe not…
> 
> Any idea of how I can troubleshoot this issue?
> Is there any log file to see the output of qemu-kvm for each instances?
> Should I use a kernel and qemu-kvm package from jessie-backports too?

Try running your instances with disk_aio=native (gnt-cluster modify -H 
kvm:disk_aio=native). This will require a hypervisor-level restart of 
all instances (i.e. gnt-instance reboot <instance>).

Regards,
Apollon