Ok it's done bugzilla 1907973, i found this message "Guest CPU doesn't match specification: missing features: tsx-ctrl (migration:294)" it's seems you're theory is good. De: "Arik Hadas" <ahadas@redhat.com> À: "Lionel Caignec" <caignec@cines.fr> Cc: "users" <users@ovirt.org> Envoyé: Mardi 15 Décembre 2020 15:50:26 Objet: Re: [ovirt-users] Re: Bad CPU TYPE after Centos 8.3 On Tue, Dec 15, 2020 at 4:41 PM Lionel Caignec < [ mailto:caignec@cines.fr | caignec@cines.fr ] > wrote: Hi Arik, Yes, my problem about cpu type, seems linked to the fact i've upgraded a host in 8.3 before ugprade ovirt in 4.3... I've done another test, for my problem of vm migration. I start a guest directly on the 8.3 nodes then it can be migrate from/to 8.2/83. I've also tried to force it to start on a 8.2 nodes then migrate to 8.3 nodes, and it also work. BQ_BEGIN So it seems a full power cycle on guest resolve issue. BQ_END Thanks for checking this - it makes sense that when the VM starts with CPU type and flags that are derived from the new cluster configuration, the issue doesn't happen anymore. BQ_BEGIN Ok for opening a bug on vdsm, could you please let me know where? BQ_END Sure - [ https://bugzilla.redhat.com/enter_bug.cgi?product=vdsm | https://bugzilla.redhat.com/enter_bug.cgi?product=vdsm ] BQ_BEGIN Thank you. De: "Arik Hadas" < [ mailto:ahadas@redhat.com | ahadas@redhat.com ] > À: "Lionel Caignec" < [ mailto:caignec@cines.fr | caignec@cines.fr ] > Cc: "Lucia Jelinkova" < [ mailto:ljelinko@redhat.com | ljelinko@redhat.com ] >, "thomas" < [ mailto:thomas@hoberg.net | thomas@hoberg.net ] >, "users" < [ mailto:users@ovirt.org | users@ovirt.org ] > Envoyé: Mardi 15 Décembre 2020 15:21:22 Objet: Re: [ovirt-users] Re: Bad CPU TYPE after Centos 8.3 On Tue, Dec 15, 2020 at 3:29 PM Lionel Caignec < [ mailto:caignec@cines.fr | caignec@cines.fr ] > wrote: BQ_BEGIN Hi thank you it's work now. All my host 8.2/8.3 are up and running. During reactivation of 8.2 host i saw the log line "updating cluster CPU..." BQ_END Good to see that it works for you now (the credit to Lucia). So the cluster configuration didn't change because not all hosts were UP when starting the 4.4.3 engine? BQ_BEGIN I tried to move guest from 8.2 host to 8.3 one but it failed. Engine.log : ERROR [org.ovirt.engine.core.vdsbroker.monitoring.VmAnalyzer] (ForkJoinPool-1-worker-27) [] Migration of VM 'pc115dev' to host ' [ http://bombur-adm.cines.fr/ | bombur-adm.cines.fr ] ' failed: VM destroyed during the startup. ERROR [org.ovirt.engine.core.vdsbroker.monitoring.VmsMonitoring] (ForkJoinPool-1-worker-19) [] Rerun VM '4fd61e41-ea48-4ba8-b78f-dba5acb87c9e'. Called from VDS ' [ http://dis-adm.cines.fr/ | dis-adm.cines.fr ] ' ERROR [org.ovirt.engine.core.dal.dbbroker.auditloghandling.AuditLogDirector] (EE-ManagedThreadFactory-engine-Thread-4) [] EVENT_ID: VM_MIGRATION_TO_SERVER_FAILED(120), Migration failed (VM: pc115dev, Source: [ http://dis-adm.cines.fr/ | dis-adm.cines.fr ] , Destination: [ http://bombur-adm.cines.fr/ | bombur-adm.cines.fr ] ). ERROR [org.ovirt.engine.core.vdsbroker.monitoring.VmAnalyzer] (ForkJoinPool-1-worker-13) [] Migration of VM 'pc115dev' to host ' [ http://bombur-adm.cines.fr/ | bombur-adm.cines.fr ] ' failed: VM destroyed during the startup. ERROR [org.ovirt.engine.core.vdsbroker.monitoring.VmsMonitoring] (EE-ManagedScheduledExecutorService-engineScheduledThreadPool-Thread-39) [] Rerun VM '4fd61e41-ea48-4ba8-b78f-dba5acb87c9e'. Called from VDS ' [ http://dis-adm.cines.fr/ | dis-adm.cines.fr ] ' ERROR [org.ovirt.engine.core.dal.dbbroker.auditloghandling.AuditLogDirector] (EE-ManagedThreadFactory-engine-Thread-38) [] EVENT_ID: VM_MIGRATION_TO_SERVER_FAILED(120), Migration failed (VM: pc115dev, Source: [ http://dis-adm.cines.fr/ | dis-adm.cines.fr ] , Destination: [ http://bombur-adm.cines.fr/ | bombur-adm.cines.fr ] ). vdsm.log (target host) : WARN (Reactor thread) [vds.dispatcher] unhandled write event (betterAsyncore:184) WARN (jsonrpc/7) [root] ping was deprecated in favor of ping2 and confirmConnectivity (API:1349) WARN (vm/4fd61e41) [virt.vm] (vmId='4fd61e41-ea48-4ba8-b78f-dba5acb87c9e') Couldn't destroy incoming VM: Domaine non trouvé : aucun domaine ne correspond à l'UUID '4fd61e41-ea48-4ba8-b78f-dba5acb87c9e' (vm:3738) BQ_END The migration probably fails due to incompatible CPU flags - please file a bug and attach VDSM and libvirt logs from both source and destination hosts. BQ_BEGIN Another think i've noticed on event log of 8.3 host : VDSM [ http://bombur-adm.cines.fr/ | bombur-adm.cines.fr ] command SpmStatusVDS failed: Cannot inquire Lease(name='SDM', path='/dev/7586152d-338c-415d-93f4-74efd09c02fa/leases', offset=1048576): (2, 'Sanlock get hosts failure', 'No such file or directory') But file seems to exist on host : ls -l /dev/7586152d-338c-415d-93f4-74efd09c02fa/leases lrwxrwxrwx. 1 root root 7 Dec 15 14:14 /dev/7586152d-338c-415d-93f4-74efd09c02fa/leases -> ../dm-7 I tried to reboot host & restart engine, set my 8.3 host as SPM ... same error. Lionel. BQ_END BQ_END