Il giorno dom 10 mag 2020 alle ore 17:00 Gianluca Cecchi <gianluca.cecchi@gmail.com> ha scritto:
On Tue, Apr 28, 2020 at 6:48 PM Sandro Bonazzola <sbonazzo@redhat.com> wrote:
Adding +Lev Veyde , +Steve Goodman , +Evgeny Slutsky and +Gobinda Das for awareness

Il giorno sab 18 apr 2020 alle ore 13:03 Gianluca Cecchi <gianluca.cecchi@gmail.com> ha scritto:
Hello,
I saw from some mails in the list that the approach for upgrading from 4.3 to 4.4 will be similar to the 3.6 => 4.0 one.
Reading the RHV 4.0 Self Hosted Engine Guide it seems it was necessary to have a second host, going through this:
5.4. Upgrading a RHEV-H-Based Self-Hosted Engine Environment
and also

upgrade process involved:

- set global ha maintenance
- install a new host (and set it as on hosted engine one)
- migrate engine VM to this host and set the host as the SPM
- create a backup with engine-backup, verify its contents and copy to the host in some directory
- run the upgrade utility on the host to update the engine VM
hosted-engine --upgrade-appliance
this will create a backup floating disk of the engine vm disk for rollback purposes and override the existing engine disk with a new one where to deploy the new engine version applying also the restore from the engine-backup
--> what are requirements for 4.4 engine VM and so free storage to have on engine storage domain?
The engine-setup will be automatically executed on the new version engine VM
- exit from global maintenance
- update remaining hosts

Is the above procedure the correct one that I can test on a 4.3.9 lab?

The 3.6 -> 4.0 flow implied that a 4.0 host could run a 3.6 engine VM
Because the backup disk operation is done by the current running engine itself.
So it should be true also for 4.3 -> 4.4 and a 4.4. host should be able to run a 4.3 engine, correct?

In case of single 4.3.9 ovirt-node-ng host HCI with Gluster, can I apply the upgrade in a similar way?
Something like:
- create engine-backup and copy over to the host
- put host into global maintenace
- shutdown all VMS, engine included
- enable 4.4. repo
- install new image-base 4.4
is this correct and I can use 4.3 and 4.4 image base versions or do I have to sort-of-scratch necessarily? because 4.3 host is based on el7 while 4.4 on el8... but perhaps on mage based system I have flexibility..?
what version of gluster will it be applied? Incompatibilities with the current 4.3.9 one (version 6.8-1.el7)
- reboot host
- exit global maintenance and wait for engine vm to come up correctly
- enter global maintenance again
- run the upgrade utility on the host

Or do I need at least temporarily a new server to use as the 4.4. host, or what will be the expected path?

Thanks,
Gianluca



No, even if rough, indications?
I see that the also the downstream beta docs set for RHV 4.4 doesn't contain an upgrade guide yet...

Gianluca


issue with single hyperconverged host is that you need to reinstall the host OS. So only way is to make a backup of it for restore purposes in case something goes wrong (can probably use clonezilla for making a disk image), backup the engine data with engine-backup and ensure to have a backup of the VMs storage too.
Fully redeploy the node in hyperconverged deployment, trying to avoid touching the VMs storage, restore the engine data and re-run engine-setup to finish the upgrade.
If storage got compromised during the redeploy you must be able to restore it from backups. Worst case you should be able to restore the full backup you took before starting the upgrade


--