[ovirt-users] Repeated inability to complete setup of hosted engine

[Yedidyah Bar David]

----- Original Message -----
> From: "Simone Tiraboschi" <stirabos at redhat.com>
> To: "Cale Bouscal" <lunix.captain at gmail.com>
> Cc: users at ovirt.org
> Sent: Wednesday, March 25, 2015 10:31:26 AM
> Subject: Re: [ovirt-users] Repeated inability to complete setup of hosted engine
> 
> 
> 
> ----- Original Message -----
> > From: "Cale Bouscal" <lunix.captain at gmail.com>
> > To: users at ovirt.org
> > Sent: Saturday, March 21, 2015 9:15:35 PM
> > Subject: [ovirt-users] Repeated inability to complete setup of hosted
> > engine
> > 
> > Hi,
> > 
> > I have tried setting this up several times and am getting stuck post-engine
> > setup, at the host setup stage. I receive an error stating that we cannot
> > connect to "the host" (I assume that is the bare metal system hosting the
> > engine guest in this context, would be nice if it could echo the command
> > it's issuing but cannot see it in the logs). DNS and ssh are working fine,
> > I
> > can ssh between the host and guest without issue. SElinux is disabled, as
> > is
> > iptables. I can't figure out how to move forward. I'm running CentOS 7 and
> > ovirt 3.5. Here is the last of the log file before I get booted out of the
> > installer:
> > 
> > http://pastebin.com/LDqFHSZw
> > 
> > So my questions are:
> > 
> > 1) What am I missing here?
> 
> The engine VM will contact your host via SSH in order to deploy it within
> oVirt; it seams that is failing there.
> Cold you please try to connect from the engine VM to the host via SSH?
> Could you please attach your host-deploy logs from the engine VM (the should
> be under /var/log/ovirt-engine/host-deploy/) ?
> 
> > 2) how do I pick up from where I left off and use the engine I've
> > successfully created?
> 
> You can try to run
> hosted-engine --deploy --config-append=/etc/ovirt-hosted-engine/answer.conf
> but it success really depends from how and where it stopped on first attempt
> (a partial failed setup it's by definition a dirty environment)
> If you get any warning or errors it would be better to redeploy the engine VM
> cleaning the shared storage.

In addition, I'd like to point out a somewhat significant change we did [1]
between 3.5.0 and 3.5.1.

Until this change, when hosted-engine --deploy asked the engine to add the
host, it told it the IP address of the bridge created over the NIC chosen
for that. So the user had basically full control over the process, knowing
which IP address will be used etc.

After the change, we ask the engine to add the host using its hostname (the
output of the command 'hostname', or more accurately, of the command:
python -c 'import socket; print socket.gethostname()'
which should be the same).

This means that when the engine machine tries to add the host, it should both
be resolvable using that name, and be accessible, i.e. the name should point
at the IP address of the interface the engine should connect to.

For now, if your dns is not working, you'll have to add this on the engine
machine by hand. E.g. if the hostname of the host is host1.example.com, and
the IP address the engine should connect to is 192.168.0.1, run on the engine
machine (and better on all hosts, including the one you install on and the
ones you'll add later - this will be needed for migrations):

echo 192.168.0.1 host1.example.com >> /etc/hosts

We have [2] to make this configurable.

[1] https://gerrit.ovirt.org/#/q/I6a2fc79fe6e57892f0cb9327da1ab8d44b15f29f,n,z
[2] https://bugzilla.redhat.com/show_bug.cgi?id=1188675

Best,
-- 
Didi