Virtual Center Redundancy/HA

[gemona]

Hello All:

Want to give you an idea of the environment we are running.

We run VC 2.5 on a single server. In our VMware environment, we have about 15 ESX 3.5 hosts and planning to move to Vsphere 4.0 that we manage through VC.

Since right now the VC is a single point of failure, I'd like to put some level of redundancy behind it, and if it is possible in a HA environment. I want to get ideas on how this is possible. Perhaps, some of you guys out there have done some work on this and implemented this in a production environment. Your insights on this would be highly appreciated.

Kind Regards

=GEM

[AndreTheGiant]

There are at least 3 solutions to have HA on VC:

• use a VM for VC and use VMware HA (in this case the downtime could be some minutes...)

• use a MSCS solution for VC: http://www.vmware.com/pdf/VC_MSCS.pdf

• use vCenter Server Heartbeat product: http://www.vmware.com/pdf/VC_MSCS.pdf

IMHO, for small/medium environment I prefer the VMware HA solution... very simple and cheap.

Be sure to have also the vCenter DB on your VM, or you have to find a HA solution also for your DB.

Andre

[krowczynski]

For your enviroment I would also prefered Andre's suggestion!

MCP, VCP

[Henk_Willemse]

I would agree as well, have VCenter hosted on one of your hosts and disable the DRS functionality for that VM.

Make sure if you use HA that you set the restart priority to the highest. You want to make sure that vCenter will reboot first as other VM's might be depending on one of its services.

If your hosting the DB on a separate VM you will have loads options there, you could build a cluster for it or us one the native DB mirrroring options in SQL (data gaurd in oracle) itself.

Remember that all information is stored in the DB so that one is the most important, you could have a 2nd vCenter VM in standby and just connect that to the DB in case of failure and continue to work.

hope this helps !!

[lmonaco]

As mentioned before, either use MSCS or just mirror the database.

[gemona]

Andre:

Thanks for the reply. So to understand your first option correctly, you are suggesting doing away with the physical server and just one VM. Both VC and the DB in one VM and VMware HA will take care of the rest. Is this right ?

-M

[gemona]

Henk:

Thanks for the reply. If I have another VC as a standby, don't I have to make some configuration changes on the ESX host before the standby VC can start managing the hosts and VMs ?

Can you point me to any KB or any other document that list the steps I have to take to operationalize the standby VC in the event the Primary VC fails ?

Thanks,

-M

[Henk_Willemse]

Gemona,

If you have a physical vCenter server and you want to keep that, you can P2V it and have it in standby mode in VMware, this will only work if the DB is on another server. When the Physical servers fails you can login to the ESX server using the vic and start/unsuspend the VM, all information is stored in the DB so this VC VM should be able to pick everything up.

give it a try, you should be able to create this scenario in your setup already.

[AndreTheGiant]

you are suggesting doing away with the physical server and just one VM. Both VC and the DB in one VM and VMware HA will take care of the rest. Is this right ?

It could be an idea...

You can make backup of your VC and also disaster recovery.

And, if you have VMware HA, you can automatic restart VC if one ESX fail.

But you need (of course) shared storage.

Andre

[gemona]

This idea didn't fly with management. They want a standby server (could be a VM with a copy of the DB) that I can manually configure in the case active physical server crashes. I am thinking of the things that this standby server needs to have to run VC

1) Latest copy of the DB. This we can do by setting up replication between the active DB and standby DB

2) Both should use the same public/private ssl keys

2) ESX hosts that were managed by the active VC needs to know about the new VC. I believe this can be done having both VC servers have the same I.

Correct me if I am wrong but I think this is doable. Am I missing anything in the steps ?

-Gem

[gemona]

Henk:

If I p2v the server, can I not have the database replicated to the new server that the standby VC is going to be. This way I have the latest copy of the database that the standby VC can point to. I am thinking below are the steps that I would need to do to have the standby server become the active VC.

1) Latest copy of the DB. This we can do by setting up replication between the active DB and standby DB

2) Both should use the same public/private ssl keys

2) ESX hosts that were managed by the active VC needs to know about the

new VC. I believe this can be done having both VC servers have the same

I.

Anything else I am missing ?

-GM

[AndreTheGiant]

Correct me if I am wrong but I think this is doable. Am I missing anything in the steps ?

It could work... but this is just what vCenter Server Heartbeat will do... consider this product that can help you in those task.

Andre

[Henk_Willemse]

Gemona,

you're right but i would advise you to have your db on a seperate server, this allows you to keep a "clone" of the vc without changing the vc details in all esx hosts.

sql can then be replicated by using the native tools or mscs.

you could do this with a vc physical and virtual and the same with sql.

[gemona]

Henk:

Just curious, where would I change the VC details on the ESX hosts ?

-GEM

[bulletprooffool]

HA continues to fnction, even if your VC goes down.

You can install a second instance of VC on a different host, point it to the same DB and stop the services running, then simply set your DNS up with an Alias that points to whichever host ios live at a given time and flip as needed.

The prettiest solution though is VMWare heartbeat (though this is horrifically expensive)