On Thu, Aug 23, 2018 at 10:19 AM Gianluca Cecchi <gianluca.cecchi@gmail.com> wrote:
On Wed, Aug 22, 2018 at 5:44 PM Simone Tiraboschi <stirabos@redhat.com> wrote:


Thanks for the answer, Simone, but I have not understood.
I used ovirt-node-ng-installer-ovirt-4.2-2018082006.iso to install the OS host node. 
Does this resemble what you call "new node zero deployment flow"?
When connecting to cockpit and selecting hosted engine, I get this screen:

what to choose and if any particular setup needs?
As I wrote, the command "gdeploy -c gdeploy.conf" completed without errors on the host and now I have the 3 (small, only to test the workflow for now) gluster volumes running

Now you can simply continue with "Hosted Engine" button in cockpit.
You could also skip the manual execution of gdeploy from CLI and do everything from cockpit with the "Hyperconverged" button but now you already did it.


Thanks for your time, Simone.
The installation went ok until the problem below due to my engine gluster volume too small (I set up 60Gb for it and it didn't fit for some MBs..). I got:
 
[ INFO ] TASK [Check storage domain free space]
[ ERROR ] fatal: [localhost]: FAILED! => {"changed": false, "msg": "Error: the target storage domain contains only 59.0GiB of available space while a minimum of 60.0GiB is required If you wish to use the current target storage domain by extending it, make sure it contains nothing before adding it."}

Then I scratched (it is a small nested test and so I simply discarded the changes) and updated gdeploy.conf to accomodate 65Gb for the engine gluster volume and all went pretty smooth then.
I also noticed that the engine VM and its storage domain is already visible in web admin portal, so there is no immediate need to add another data domain and get hosted engine imported as before, correct?

Yes, as expected with the new flow.
Theoretically you can just end with a single SD for hosted-engine and regular VMs but keep them distinct is highly recommended for maitenability.  
 

I'm up and running and I'm going to scratch again and test your suggestion to bypass also the gdeploy.conf step and go all with gui.
In fact I notice that when configuring storage there is a note saying 
"
Please note that only replica 1 and replica 3 volumes are supported.
"

.. I'll let you know if it goes ok the all gui setup.

In the mean time one further question: it seems that the cluster configured has not the "gluster service" checkbox enabled, so I think it is for that reason that I don't see any gluster related configuration information inside the web admin portal....
Is this expected with this approach?
Is something different expected when trying instead the full gui approach? Will the web admin portal aware of the gluster config in that case?

Yes, in that case hosted-engine-setup will be aware of the hyperconverged configuration and it will automatically configure the cluster for the gluster server adding also the two additional hosts.
 

Also because a further step I would like to try is to extend this initial setup to a 3-node based gluster one...

Thanks again,
Gianluca