[ovirt-users] Frustration and Contradictory Guidance?

[Simone Tiraboschi]

On Fri, Dec 18, 2015 at 9:37 PM, Eriks Goodwin <eriks at connectcombpo.com>
wrote:

> Hi all,
>
> I have been working on building a new oVirt 3.6 deployment for more than
> two weeks now and a multitude of things are frustrating my attempts...
> Would someone please point me to the "right" HOWTO article for deploying a
> production-grade oVirt 3.6.x system with HA hosted engine?
>

Did you tried following this:
http://www.ovirt.org/Features/HEApplianceFlow

Or this:
https://www.youtube.com/watch?v=ODJ_UO7U1WQ


>
> I spent almost two weeks just trying to get the networking straight--only
> to have VDSM overwrite all my hard work and totally break my networking (I
> was using a bonded set of NICs with bridges and VLANs for ovirt-management,
> ovirt-storage, and local-lan.).
>
> Once I abandoned VLANs and installed a third physical NIC, I was able to
> install the first node and the hosted engine-- but after I got things
> "working" now I receive constantly cycling email messages for
> ovirt-hosted-engine state transition: ReinitializeFSM-EngineStarting,
> EngineStarting-EngineUp, and StartState-ReinitializeFSM
>
>
Involving Martin here.


> When I try to follow the instructions to add engines for HA, the system
> keep faulting that it is unable to find the answer file it needs on node 2.
>

The oVirt engine runs on a VM, than you need al least two physical hosts
and and some kind of shared storage (NFS, iSCSI, Gluster, FC).

You should run hosted-engine --deploy on the first host, it configures it
and starts the engine VM here.

Then you should run hosted-engine --deploy also the the second host
pointing to the same shared storage. The setup will find the existing
storage domain and it asks is you want to add an host, you have to say yes,
this is host 2.
At this point the setup will ask you the address of the first host to
download an answerfile from there to configure the second host as the first
without having you to manually enter the same data.


>
> And now that I finally figured out that I needed to add "another" storage
> domain for the datacenter to initialize,
>

The engine VM is on its own dedicated storage domain, you need an
additional one for other VMs.


> the addition of new hosted-engine deployments is blocked by the presence
> of the attached storage domain.
>

?
sorry, can you better explain the issue or add a log?


>
> So, all of that said....  is there a HOWTO written specifically for CentOS
> 7 that does not want me to kill Network Manager and switch from FirewallD
> back to iptables--and does not assume I want to use glusterfs?
>

GlusterFS is just an option. iptables instead of Firewalld and no
NetworkManager are mandatory.


>
> I have two sets of HP Proliant servers I want to use for this:  4 DL380
> G6s and 4 DL380G5s, plus one G5 which is my initial NFS3 share
>
> For such an AWESOME project, I find it really disappointing that the
> documentation is so poor/outdated.  :-(
>
> Thanks,
> Eriks
>
>
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20151218/c2cfed1b/attachment-0001.html>