RHEV-M and RHEV-H on same Physical Server

Latest response

Is it possible to run RHEV-M and RHEV-H on same physical Server.


Hi Alok,


At the moment it is not possible or supported. Running RHEV-M on a hypervsor it manages is a future feature





What would stop this from being possible?  As long as I can create a VM in a rhev-h physical machine, I should be able to install rhev-m and add the hypervisor host to it?


This means I can't use rhev at all unless I have 2 physical machines?


Justin, first of all, you won't be able to do that. If you ever installed RHEV, you know that after you have RHEV-M running, you still need to connect a host to it, and then through this host, define the storage domains.

Once you have storage domains ready, then you can set up VMs. So without RHEV-M, a host and storage, you cannot add VMs. 


This is just one among several reasons what you want to do will not be supported, or possible without some hacking through the entire system.


Just keep in mind that while we are working on making this possible eventually, (because there are lots of requests for this) RHEV is an enterprise grade system, built and designed to manage huge hypervisor clusters, made up of hundreds of hosts, running thousands of VMs, and single standalone hosts are not in the main scope, when scalability is the main focus. 

Still, if this is very important for you, please feel free to open a support case, and request a feature there - support will attach your request to the existing feature request, which will lend it more weight when it is considered for inclusion in the upcoming versions

hi, thanks for the reply. 


Ok, then would it be possible to use a temporary rhev-m host to create the first VM and storage domain, andthen install rhev-m inside the VM, and reconnect the rhev-h machine to the newly created rhev-m VM?


Its a shame because in VMware you can just use the vsphere client to create the first VM, andthen install vcenter on that.  Rhev doesn't seem as flexible this way, and i'd much rather use rhev..




What you suggest will present other issues. As I have alreayd mentioned, we are working on a way to make this possible, but it will not be fast or easy, and we really don't like delivering half-baked solutions.


If you are looking for a way to cut down the host count, for now this cannot be helped. If you are looking to provide high availability for RHEV-M, there are several different ways to do that

Can you say what these "other issues" would actually be?  I was tempted to try it just to see, but maybe you have tried doing this already?

It works in theory, and of course no proper QA was done on an unsupported solution. However, since RHEV-M manages the resource monitoring for HA, what will happen if the reource the RHEV-M VM is on dies?


There is a different approach, it is not readily available out of the box, and is not supported without a GPS engagement, but you could have a RHEV-M cluster on baremetal, and use the same hosts as hypervisors as well. Again, for this sort of thing, you'd need to work with our professional services in order to have it working, and what's more important - to have it supported.


I really think though, that this is rather redundant - if you have a very large and critical setup, you probably can spare a couple of hosts to build an HA RHEV-M cluster. And if the setup is very small, and you can't afford an extra box/blade in there, having a good backup of RHEV-M is usually enough - when the manager is down, the hosts keep working and the VMs are not stopped, so if you bring RHEV-M back up in a reasonable amount of time, no harm done

Ovirt upstream calls the engine and hypervisor on the same host 'all-in-one' and there is a plugin for Fedora 18 that enables.

It works well, but could be taxing on the host, but is useful for testing.

package name from Fedora 18:  ovirt-engine-setup-plugin-allinone-3.1.0-2.fc18.noarch.rpm

Since RHEL is based on Fedora, I would hope that this makes its back to RHEV.

Here you can find more info about this in RHEV:


RHEV 3.1 comes with a rhevm-alinone-plugin package, that sets up a host with RHEV-M, local storage and vdsm. This host can be attached to another host cluster, and you will end up with a RHEV-M + vdsm host, and a bunch of other hosts. 

All-in-one is not supported for production though, it is there as a means for demoing RHEV on a standalone machine, and can be useful for a presales person out in the field (that's really the targeted usecase)

The above is not production use case. It's only for testing and demo purpose.

With 3.3, you can have a self-hosted engine

Thanks for adding this here, Sergio