[ClusterLabs] [EXT] moving VM live fails?

[lejeczek]

On 24/11/2023 08:33, Windl, Ulrich wrote:
> Hi!
>
> So you have different CPUs in the cluster? We once had a similar situation with Xen using live migration: Migration failed, and the cluster "wasn't that smart" handling the situation. The solution was (with the help of support) to add some CPU flags masking in the VM configuration so that the newer CPU features were not used.
> Before migration worked from the older to the newer CPU and back if the VM had been started on the older CPU initially, but when it had been started on the newer CPU it could not migrate to the older one. I vaguely remember the details but the hardware was a ProLiant 380 G7 vs. 380 G8 or so.
> My recommendation is: If you want to size a cluster, but all the nodes you need, because if you want to add another node maybe in two years or so, the older model may not be available any more. IT lifetime is short.
>
> Regards,
> Ulrich
>
> -----Original Message-----
> From: Users <users-bounces at clusterlabs.org> On Behalf Of lejeczek via Users
> Sent: Friday, November 17, 2023 12:55 PM
> To: users at clusterlabs.org
> Cc: lejeczek <peljasz at yahoo.co.uk>
> Subject: [EXT] [ClusterLabs] moving VM live fails?
>
> Hi guys.
>
> I have a resource which when asked to 'move' then it fails with:
>
>   virtqemud[3405456]: operation failed: guest CPU doesn't match specification: missing features: xsave
>
> but VM domain does not require (nor disable) the feature:
>
>    <cpu mode='custom' match='exact' check='partial'>
>      <feature policy='require' .....
>
> what even more interesting, _virsh_ migrate does the migration live a ok.
>
> I'm on Centos 9 with packages up to day.
>
> All thoughts much appreciated.
> many thanks, L.
>
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users
>
> ClusterLabs home: https://www.clusterlabs.org/
This must go down - same environment - all the way, to the 
kernel. I knew that but forgotten - one box did change 
default kernel & after reboot booted with Centos's default 
as opposed to other boxes running with ver. 6.x.
So, my bad, not really an issue.
Thanks,. L