10:10 a.m.
On Mon, 26 Feb 2024, ovirt(a)the-dawg.net wrote:
I have multiple aging clusters that I now need to upgrade both the
host OS and the oVirt release. I have seen the notes about upgrading 4.2 to 4.3 and then
being able to upgrade to 4.5, but what's not so clear is how to upgrade the host
nodes' OS versions. After thinking about it, I wonder if this method is sane. Any
advice or guidance is greatly appreciated.
1) spin up a new cluster running oVirt 4.5.x on the latest Rocky Linux release.
2) shutdown VMs on the original cluster
3) detach the domain storage
4) snapshot the domain storage (for rollback)
5) import the domain on the new cluster.
6) reload the OS on the original cluster to the new OS release (no local storage)
7) add those hosts to the new cluster
Last year I have upgraded 2 clusters from oVirt 4.3.x (CentOS 7.x) to 4.4.x (AlmaLinux
8.x) and next to 4.5.x.
I have installed a new standalone manager using AlmaLinux 8.x following the
documentation (Extract from my notes):
- enabled the following modules: javapackages-tools, pki-deps, postgresql:12,
mod_auth_openidc:2.3, nodejs:14
- installed
- nfs-utils
- iscsi-initiator-utils
- lsscsi
- ovirt-engine
- ovirt-engine-extension-aaa-ldap
- ovirt-engine-extension-aaa-ldap-setup
Copy /etc/ovirt-engine/aaa/*.properties so it will be possible to use
LDAP to authenticate
Set /etc/iscsi/initiatorname.iscsi to match the current manager
On the old standalone manager (CentOS 7.x)
cd /srv/backup/
engine-backup --scope=all --mode=backup --file=backup.bck --log=backuplog.log
scp -p backup.bck new-manager:/srv/backup/
hostnamectl set-hostname old-manager
nmcli to change the the IP
reboot
On the new manager
- nmcli to use the IP that was used previsouly on the manager
- reboot
- Restore from backup.bck file
cd /srv/backup/
engine-backup --mode=restore --file=backup.bck --provision-all-databases
-
export OTOPI_DNF_ENABLE=1
engine-setup
I had to downgrade a package
wget -N
https://vault.centos.org/centos/8/AppStream/x86_64/os/Packages/postgresql...
rpm -Uvh --oldpackage postgresql-jdbc-42.2.3-3.el8_2.noarch.rpm
Finaly the new manager was working correctly.
It was time to replace each node one by one:
- Cluster / "Stratégie d'ordonnancement" /, replace vm_evenly_distributed by
none to prevent migration
- put a node in maintenance
- reinstall the node in AlmaLinux 8.x
- deploy oVirt 4.4.x on it
- check/configure the network
- activate the node
- shutdown a VM and restart it on the new node
Next step has been to update the cluster level (and reboot each VM) and it has been
possible
to update to 4.5 after that.
I had to update the firewall rules on the manager, something like
cp -p /etc/ovirt-engine/firewalld/ovirt-nfs.xml /etc/firewalld/services
firewall-cmd --permanent --add-service ovirt-nfs
firewall-cmd --permanent --add-service=mountd
firewall-cmd --reload
Good luck,
Christophe
--
,-~~-.___. ._.
/ | ' \ | |--------. Christophe GRENIER
( ) 0 | | | grenier(a)cgsecurity.org
\_/-, ,----' | | |
==== !_!-v---v--.
/ \-'~; .--------. TestDisk & PhotoRec
/ __/~| ._-""|| | Data Recovery
=( _____|_|____||________| https://www.cgsecurity.org