SRM Licensing Confusion

[TheKubMI]

I apologize if this has been answered before but I'm a bit confused and I don't know if I have the correct info. We have two datacenters each with hosts and a SAN (4 hosts in primary, 3 in DR site) and SRM configured. All VMs run in the primary datacenter, so at any given time only one is running "protected" machines.

I was going through the SRM documentation when I came across this statement:

A protection enablement license key (SRM_PROTECTED_HOST) that specifies the number of ESX CPUs that can run protected virtual machines at a site. Install this key at the protected site to enable failover. Install it at the recovery and protected sites to enable bi-directional operation (failback).

Obviously the hardware at the primary is better so should we have an SRM worthy failure we would obviously want to be able to FAILBACK to the primary site when it is back up and running. So my confusion is, is this telling me to copy the SAME set of licenses (in this case 8 CPU licenses) to both VC\SRM servers? Or because we would want to failback we would have to purchase additional licenses (14 CPU total) and apply the 6 additional to the DR Site?

I want to say that we would just use\transfer the licenses in the event of a failure for the failback procedure. But if we were running VMs in a split fashion (half at each datacenter) where we could fail over from either direction then we would need to license all CPUs in both data centers.

Thanks

[Saadat]

I am not completely sure about this, so someone who has SRM or a VMware employee can confirm this.

I think when you protect your VMs in your main site, you have to apply the licenses for the hosts (ESX Servers /per socket).

So you just need to license those Servers.

Now in a failback scenario: (From your DR to your Main Site) as far as I know this is not a fully automated process, so here is what happens:

You have to change the role, meaning your DR site will play the role of your main office, So DR site will be your protected site and your main office will play the DR Site role.

So license wise you should be fine, because you don't have bidirectional replication.

Till you completely fail back to your original Main office, and restart the process. (Change the roles to the original)

I hope I haven't confused you.

Ernest

[jbock]

I agree that the line in the documentation that you refer to should be clearer, I'll have it clarified in the next revision of the documentation.

The succinct summary is that you license the processors that are running virtual machines that are protected by SRM, and that to automate failback you can transfer over the licenses that were originally at the production site to the recovery site (since they're not in use at the production site at that time). You shouldn't install the same licenses in two places since that is effectively using twice as many licenses as you purchased. Whether you need to do that or not depends of course on whether you had bidirectional protection to start with and whether you need to license processors for failback that are not already licensed for failover.

Specific to your scenario, you split the total quantity of licenses you purchase across the two sites (i.e. generate two separate license files in the license portal that add up to the total number of licenses purchased and install one of those at each site). If you have to do a failover and then want to configure failback (and need more SRM processor licenses than you have at the recovery site), you can go back to the license portal and generate a license file with all of your processor licenses in a single file, install that at the recovery site, configure and perform failback, and then generate two separate license files again as before to enable you to reprotect at your rebuilt site.

Hope that clarifies,

Jon

[TheKubMI]

OK.

The succinct summary is that you license the processors that are running virtual machines that are protected by SRM, and that to automate failback you can transfer over the licenses that were originally at the production site to the recovery site (since they're not in use at the production site at that time).

Which goes along with what I was thinking. What is kinda throwing me off is...

Specific to your scenario, you split the total quantity of licenses you purchase across the two sites (i.e. generate two separate license files in the license portal that add up to the total number of licenses purchased and install one of those at each site). If you have to do a failover and then want to configure failback (and need more SRM processor licenses than you have at the recovery site), you can go back to the license portal and generate a license file with all of your processor licenses in a single file, install that at the recovery site, configure and perform failback, and then generate two separate license files again as before to enable you to reprotect at your rebuilt site.

If we are only ever going to be running the protected machines in 1 datacenter at any given time why do we split the quantity? So either you misread my original post or I am misreading yours (or the topic made my brain melt which is a likely scenario).

-

---

I also just want to take a second and thank VMWare for making a tool that makes an entire datacenter failover basicly a one click operation. However, that one click comes with a lot of additional steps.

Our GO BOOK contains has one page describing the steps needed to perform the actual recovery. :smileygrin:

But it also contains 5 pages describing the license juggleing procedure of signing in to the site, locating the right page, deleting the old licens file(s), creating a new one, downloading and saving to the correct place, opening the license manager tool, rereading the file restarting the service, rechecking the licensing status. :smileyangry:

[jbock · Accepted solution]

Sorry, should have been more careful in responding. The part about splitting the license file is about your scenario "...if we were running VMs in a split fashion (half at each datacenter)

where we could fail over from either direction then we would need to

license all CPUs in both data centers." Apologies for missing the fact that there is an "if" in that.

If you are only running protected VMs at one site you wouldn't have to split the license file, as you say. You'd install it at the production side, then after failover you could install the same license file at the recovery side to enable failback.

--Jon

[TheKubMI]

Thank you for clarifying.