On Wed, May 5, 2021 at 8:47 PM Derek Atkins <derek(a)ihtfp.com> wrote:
On Wed, May 5, 2021 9:24 am, Yedidyah Bar David wrote:
> On Wed, May 5, 2021 at 4:20 PM Derek Atkins <derek(a)ihtfp.com> wrote:
>> On Wed, May 5, 2021 1:40 am, Yedidyah Bar David wrote:
>> >> And how about engine storage domain? I have a new NFS mount ready but
>> >> did not ask me about which storage domain to use. Will it ask in
>> >> stage?
>> > Yes, storage is handled (also asked about) in a later stage, after the
>> > engine is up and the host added to it.. This is a change from <= 4.2,
>> > and was done so that we can use the engine (and vdsm) for this,
>> > instead of duplicating their functionality in the deploy code.
>> Color me confused... How does this work?
> Please review:
Thanks. However nowhere in those slides does it talk about when one would
install the Engine VM base OS. So my questions (quoted below) still
You are most welcome to simply study the (ansible) sources:
Basically, the process:
1. Prompts for various details
2. Creates a local libvirt VM on local storage from the appliance
image (+these details)
3. Starts this VM as HostedEngineLocal
4. Runs engine-setup, adds the host to the engine
5. Then prompts for storage details etc.
6. Creates the storage domain, edits the engine db a bit, shuts down
the VM, edits it a bit further
7. Copies the VM's disk to the shared storage
8. Starts the HA daemons and let them start the engine vm
I hope this clarifies.
>> If I'm installing CentOS and then the Engine into the Hosted-Engine VM,
>> don't I need a storage domain for that OS defined in order for me to
>> install the OS into the VM? This sounds like a bootstrapping problem?
>> Or do I install into some temporary storage and then once the Engine +
>> VDSM are up and running I can assign the permanent storage and it will
>> save/copy the temporary storage over to the permanent storage? But in
>> the latter case, that would imply the ability to move the hosted-engine
>> storage domain from one place to another?
See above. In theory, you can indeed manually update conf where needed to
move the engine to another storage. In practice this was never considered
important enough, and somewhat risky, so was never done.
The documented way to do that is using engine-backup and hosted-engine
>> Or does 4.4 remove the ability to install the hosted-engine into a
>> OS and instead one must always use a Hosted Engine appliance?
This was done in 4.1, 4 years ago:
>> NB: I installed back at 4.0 and have been upgrading ever since, so I've
>> never had to re-install the engine from scratch; I know I will need to
>> that if I ever decide to upgrade from 4.3.10 to 4.4 (which, frankly, I
>> not looking forward to doing any time soon).
So this is indeed new to you. I hope that's ok :-).
At first I was also somewhat hesitant about this, losing so much flexibility.
But in practice, I can't remember even one case of someone asking to have
this back. The appliance simply works, and works well, is far easier and
faster to install than anything else (unless you had a very fine-tuned
PXE installation system), and we still provide enough flexibility (IMO) using
ansible hooks - see above link to the sources (github shows you the README
in that directory).
And, BTW, the appliance is not like ovirt-node - once you install it, it's
just a plain normal CentOS machine - you upgrade stuff using 'dnf update'
and 'engine-setup', etc.
>> > Good luck and best regards,
Derek Atkins 617-623-3745
Computer and Internet Security Consultant