Hi Lionel,

in oVirt 4.4.2 the Secure Intel Cascadelake Server Family was configured to use the Cascadelake-Server libvirt/qemu CPU. In oVirt 4.4.3, this has been changed to Cascadelake-Server-noTSX libvirt/qemu CPU (since the TSX was dropped in 8.3).

The problem you're facing is caused by this transition. In the cluster table you can see that the running hosts still use the old configuration - Cascadelake-Server. That is why the new 8.3 host is non operative.

Now the question is why the configuration in the cluster table was not updated. This should be the correct workflow:
1. engine update from 4.4.2 to 4.4.3
2. during engine-setup, the configuration values should be updated in the vdc_table in the database
3. during engine startup, the cluster value should be updated IF all hosts are UP AND all hosts comply to the new configuration

Could anything from the below cause the configuration in the cluster table not to be updated?
a) engine-setup was not performed
b) not all hosts in the cluster were UP
c) some of the 8.2 hosts do not support the Cascadelake-Server-noTSX (I can see that the 8.2 you've checked does support it, maybe it would be worth checking the others as well).

Based on your answer we can try a workaround in the Webadmin before changing the kernel values.

Regards,

Lucia

On Tue, Dec 15, 2020 at 11:01 AM Lionel Caignec <caignec@cines.fr> wrote:
Hi All,
and thanks for helping me

So if i does not mistakte, we cannot use anymore cascadelake Server in centos 8.3?
Cluster configuration is Secure CascadeLake Server Family



Here is output of the sql query :

    name    |                cpu_name                |                  cpu_flags                   |                                  cpu_verb                                   
------------+----------------------------------------+----------------------------------------------+-----------------------------------------------------------------------------
 moria | Secure Intel Cascadelake Server Family | vmx,mds-no,md-clear,model_Cascadelake-Server | Cascadelake-Server,+md-clear,+mds-no,-hle,-rtm,+tsx-ctrl,+arch-capabilities


So in your link they said TSX is disabled by default, may i enable it? or the best option is to change cluster cpu type to lower value?

Sorry for all my question i'm quite lost


Regards,
Lionel.
De: "Lucia Jelinkova" <ljelinko@redhat.com>
À: thomas@hoberg.net
Cc: "users" <users@ovirt.org>
Envoyé: Mardi 15 Décembre 2020 10:20:08
Objet: [ovirt-users] Re: Bad CPU TYPE after Centos 8.3

Hi Lionel,

Thank you for the output from the virsh command. The interesting part is this:
<model usable='yes'>Cascadelake-Server-noTSX</model>
<model usable='no'>Cascadelake-Server</model>
The Cascadelake-Server is not usable any more in 8.3 [1] yet that is what the cluster is probably configured with. 
Could you please run the following query for your cluster to be sure?

select name, cpu_name, cpu_flags, cpu_verb from cluster;

Could you please also run the virsh domcapabilities command on one of your active and running 8.2 hosts to see if they support Cascadelake-Server-noTSX? 

1: https://www.phoronix.com/scan.php?page=news_item&px=Red-Hat-RHEL-8.3

Thank you, 

Lucia






_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-leave@ovirt.org
Privacy Statement: https://www.ovirt.org/privacy-policy.html
oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/
List Archives: https://lists.ovirt.org/archives/list/users@ovirt.org/message/PWYZH3SK2WALLMWYS45EDAENP4IJQLSZ/