Re: Using CentOS-6.x on KVM-hosts - what are the threats?

[ilya <il...@gmail.com>]

Vladimir

It comes down to version of qemu and cpu flags. As you know this is not
related to cloudstack.

Because of major differences in qemu and cpu flags -  safest method is
power down and power up.

QEMUs claim to fame was that it can migrate from AMD to Intel and vice
versa. In practice its hit or miss.

What is guest.cpu.mode set to in your agent.properties?

Also, this maybe a good help for "guest.cpu.mode" capabilities.

https://libvirt.org/formatdomain.html#elementsCPU

Regards
ilya

On 2/25/17 2:29 AM, Vladimir Melnik wrote:
> Thank you for the comment, Simon! The most funny thing is that I've added 3
> new hosts to my infrastructure in December, their hardware is awesome, so
> their hardware refresh is not a matter of the nearest future. :) Anyhow, 6.x
> works great and the only thing I regret is lacking certain features (such as
> IOpS limits).
> 
>  
> 
> I'm also maintaining a cluster of 5 hosts (primary storages aren't local,
> they're connected via GlusterFS & NFS) which've been running 6.x too. I've
> upgraded 3 of hosts from 6.x to 7.x, but when I'm trying to migrate a VM
> from the "old" hosts to the "new" ones, the migration is being timed out and
> the VM is being frozen in the "paused" state. I noticed a difference in the
> CPU-flags set: all the hosts running 7.x have the "nopl" flag, but the hosts
> running 6.x don't. This option appears only after installing 7.x and maybe
> this is the cause. Does anyone have any suggestions on the reason that
> causes freezing the VMs when they've been migrating from the 6.x-powered
> hosts to the 7.x-powered ones? Is that the "nopl" flag? Is that anything
> else? Thanks to all!
> 
>