Hello,

On Thu, Dec 24, 2020 at 12:45 PM Strahil Nikolov <hunter86_bg@yahoo.com> wrote:

In the WebUI, I see all 3 hosts marked as "up".
But when I run hosted-engine --vm-status (or migrate the hosted engine), only the first (original deployed) host is available.
I tried "reinstalling" (from the WebUI) the two hosts, no errors, no change.
I tried upgrading the cluster again, host 2 / 3 (the "missing" hosts) upgrade is successful; hosts1 fails (cannot migrate the hosted engine).

Start debugging from the hosts that are not seen via "hosted-engine --vm-status".
There are 2 log files to check: /var/log/ovirt-hosted-engine-ha/{broker,agent}.log

Just check what they are complaining about.
Also , you need both ovirt-ha-broker.service and ovirt-ha-agent.service up and running. If there is an issue, the ovirt-ha-agent is being restarted (so it's expected behaviour).

Usually the broker should stop complaining and then the agent will kick in.

Best Regards,
Strahil Nikolov

I "forked" this email into a new subject ("Manually deploying a hyperconverged setup with an existing gluster bricks")

But nevertheless, no idea if its intended or not, but it seems that adding a host via "host->computer->create" doesn't create the necessary /etc/ovirt-hosted-engine/hosted-engine.conf configuration on the new hosts, preventing ovirt-ha-* services from starting.

Thanks,

Gilboa