[ovirt-users] oVirt non-self-hosted HA

[Johan Bernhardsson]

The norm is to have a cluster with shared storage. So you have 3 to 5 
hardware noed that shares storage for the hosted engine. That shared 
storage is in sync. So you don't have one engine per physical node.

If one hardware node goes down the engine is restarted on another node with 
the help of hosted-engine service.

/Johan

On April 5, 2018 08:11:50 TomK <tomkcpr at mdevsys.com> wrote:

On 4/4/2018 3:11 AM, Yaniv Kaul wrote:


On Wed, Apr 4, 2018 at 12:39 AM, Tom <tk at mdevsys.com
<mailto:tk at mdevsys.com>> wrote:



Sent from my iPhone

On Apr 3, 2018, at 9:32 AM, Yaniv Kaul <ykaul at redhat.com
<mailto:ykaul at redhat.com>> wrote:



On Tue, Apr 3, 2018 at 3:12 PM, TomK <tomkcpr at mdevsys.com
<mailto:tomkcpr at mdevsys.com>> wrote:

Hey Guy's,

If I'm looking to setup the oVirt engine in an HA
configuration off the physical servers hosting my VM's (non
self hosted), what are my options here?

I want to setup two to four active oVirt engine instances
elsewhere and handle the HA via something like haproxy /
keepalived to keep the entire experience seamless to the user.


You will need to set up the oVirt engine service as well as the PG
database (and ovirt-engine-dwhd service and any other service we
run next to the engine) as highly available module.
In pacemaker[1], for example.
You'll need to ensure configuration is also sync'ed between nodes,
etc.
Y.

So already have one ovirt engine setup separately on a vm that
manages two remote physical hosts.  So familiar with the single host
approach which I would simply replicate.  At least that’s the idea
anyway.  Could you please expand a bit on the highly available
module and  syncing the config between hosts?


That's a different strategy, which is also legit - you treat this VM as
a highly available resource. Now you do not need to sync the config -
just the VM disk and config.

I think there's a postgres component too and if oVirt engine keeps all
it's date on the postgres tables, then synchronizing this piece might be
all I need?  I'm not sure how the separate oVirt engines sitting on
various separate physical hosts keep their settings in sync about the
rest of the physicals in an oVirt environment. (Assume we may have 100
oVirt physicals for example.)

Perhaps something like
https://www.unixarena.com/2015/12/rhel-7-pacemaker-configuring-ha-kvm-guest.html
.

But if you are already doing that, I'm not sure why you'd prefer this
over hosted-engine setup.

I'm comparing both options.  I really don't want to ask too many
specific until I have the chance to read into the details of both.

Y.

Cheers,
Tom



Cheers,
Tom


[1] https://clusterlabs.org/quickstart-redhat.html
<https://clusterlabs.org/quickstart-redhat.html>


>From what I've seen in oVirt, that seems to be possible
without the two oVirt engines even knowing each other's
existence but is it something anyone has ever done?  Any
recommendations in this case?

Having settings replicated would be a bonus but I would be
comfortable if they weren't and I handle that myself.

--
Cheers,
Tom K.
-------------------------------------------------------------------------------------

Living on earth is expensive, but it includes a free trip
around the sun.

_______________________________________________
Users mailing list
Users at ovirt.org <mailto:Users at ovirt.org>
http://lists.ovirt.org/mailman/listinfo/users
<http://lists.ovirt.org/mailman/listinfo/users>


--
Cheers,
Tom K.
-------------------------------------------------------------------------------------

Living on earth is expensive, but it includes a free trip around the sun.

_______________________________________________
Users mailing list
Users at ovirt.org
http://lists.ovirt.org/mailman/listinfo/users