[ovirt-users] Self hosted single server network requirements

[Mark Steckel]

Follow-up - I finally discovered the source of my networking troubles.

The provider (Hetzner) configures the primary IP in a somewhat odd manner for Cent OS 7.2.

For instance, here is /etc/sysconfig/network-scripts/ifcfg-eth0 (lightly edited)

DEVICE=eth0
ONBOOT=yes
BOOTPROTO=none
IPADDR=A.B.4.9
NETMASK=255.255.255.255
SCOPE="peer A.B.4.1"

And /etc/sysconfig/network-scripts/route-eth0

ADDRESS0=0.0.0.0
NETMASK0=0.0.0.0
GATEWAY0=A.B.4.1

The key thing is the SCOPE line.

This creates a private link between the server's IP and the gw IP. The server's IP address no longer has a netmask, which is now specified on the gw IP.

The odd network config appears to confuse the hosted-engine deploy process because it was unable to configure the ovirtmgmt bridge in a working manner.

Once I delete route-eth0 and changed ifcfg-eth0 to a more typical configuration all started working.

Now I just need to figure out how to configure the /29 for the VMs.

Thanks
Mark


----- Derek Atkins <derek at ihtfp.com> wrote:
> Hi,
> 
> On Mon, November 21, 2016 4:30 pm, Mark Steckel wrote:
> >
> [snip]
> >> >> > Advice so far seems to be:
> >> >> > * Use 'screen' when deploying. Easy
> >> >> > * Don't use/disable Network-Manager. Easy
> >> >> > * Preconfigure the ovirtmgmt bridge. I've got questions...
> >> >> >
> >> >> > The server has a public /32 as the primary IP, and a public /29
> >> which
> >> >> > will be used for the VMs.
> >> >> >
> >> >> > Besides creating the ovirtmgmt bridge is there anything specific
> >> for
> >> >> > how I should configure it?
> [snip]
> >
> > Yeah, I was hoping that would be the case for me too. :-)
> 
> For what it's worth, it took me about 4 tries to get it all working.
> I wound up using a script I found to clean the host and re-install it all.
>  Of course the cleanup process let the firewall in a state where it
> blocked all traffic (including SSH), but I figured that part out on the
> 2nd try so added it back into my script for the 3rd.  ;)
> 
> The other difference is that I had already created a bond0 interface which
> was my default (with a non-NM-controlled static network).  However, that
> shouldn't have made a difference.
> 
> >> What did your /etc/sysconfig/network-scripts/ifcfg-<dev> file look like?
> >> (And what does the ifcfg-ovirtmgmt file look like)?
> >
> > Sadly, thinking I messed things up I scrubbed the machine and started
> > over. I have a fresh CentOS 7.2 install waiting to run 'hosted-engine
> > --deploy' on it once I have a better sense what if anything I need to
> > prepare from a networking stand point.
> 
> Okay, so go and try it!  :)
> 
> Worst case, you need to run hosted-engine-cleanup.sh and then reset the
> firewall and reboot, and then re-install everything:
> 
> https://access.redhat.com/documentation/en/red-hat-virtualization/4.0/paged/self-hosted-engine-guide/chapter-2-deploying-self-hosted-engine
> http://www.ovirt.org/documentation/how-to/hosted-engine/#fresh-install
> 
> [snip]
> >> >> As for the /29 -- don't worry about it now, that would be a routing
> >> >> issue you can apply later.  Is it the same network as the /32?  Or is
> >> it
> >> >> a different network?
> >> >
> >> > Different.
> >>
> >> I assume both networks are available on your host interface?
> >
> > At this point only the /32 is on the host interface. The /29 is not
> > assigned at the moment.
> 
> You shouldn't need to "assign" anything in the /29 to the host.  All
> that's important is that the physical network can reach the /29.  I.e.,
> you could do something like:
> 
>      /32           /32        /29
> router ----------- host ----  VMs
>      /29
> 
> The host will bridge the network to the VMs, but it can be on a different
> network.
> 
> > Mark
> 
> -derek
> 
> -- 
>        Derek Atkins                 617-623-3745
>        derek at ihtfp.com             www.ihtfp.com
>        Computer and Internet Security Consultant
>