Create DRS group for hosts in 2 clusters

[Biga_b]

Because of the Microsoft SQL licensing per core model we need to create DRS groups for our SQL servers. We have 8 host, but need to need to run the SQL VM's on 2 host

We are in the process of migrating from ESX 4.0 to ESXi5.5. We already have vCenter 5.5 server running

We don't have the ESXi 5.5 hosts setup yet, so we must create the DRS groups on the 4,0 host. MS won't wait for us to migrate first.

We have two clusters ( 2 external data centre). Each with 4 esx 4.0 servers. Simple question:

Can DRS group only be created within one cluster? Can I create DRS groups on datacenter level. Use one host in Cluster1 and a second host in cluster2.

Thanks

[MKguy · Accepted solution]

Can DRS group only be created within one cluster? Can I create DRS groups on datacenter level. Use one host in Cluster1 and a second host in cluster2.

DRS groups can only be created in a single cluster and not across clusters. Anything else wouldn't make any sense anyways. DRS groups are for restricting automatic migrations by DRS, and DRS itself is only a cluster-wide process in the first place. So what would be the point of a single "DRS group" across clusters?

But of course you can create two separate DRS groups in your 2 separate clusters.

[Biga_b]

Thanks for the answer MKguy..

I have one last questions:   

Using the option "must run on hosts in group" will disable/ignore HA functionality if I read the articles correct. What will happen with the VM's if one of the hosts in the DR group crashes.?

They won't start up on the other host in the DR group. So should I use the option" should run on hosts in group". Sorry for the dummy question. I think I know the answer, but just looking for confirmation.

Thanks

[MKguy]

Using the option "must run on hosts in group" will disable/ignore HA functionality if I read the articles correct. What will happen with the VM's if one of the hosts in the DR group crashes.?

HA will respect MUST rules. If there is no other alive host in the according DRS host group, the VM will not be restarted on any of the other hosts.

With SHOULD rules on the other hand, they will be restarted on other hosts as well:

Mandatory DRS Rules and HA - Yellow Bricks

(Storage) DRS (anti-) affinity rule types and HA interoperability

Should or Must VM-Host affinity rules?

[Biga_b]

One thing is still not clear to me.

I want to have a active / passive situation in a DR group, otherwise I'll have to license both ESX host according to Microsoft.

With the MUST rule, will DRS still kick in, and load balance resources within those 2 ESX host if one of the hosts is leading to 100% CPU or Memory utilization?

Thanx

[MKguy]

With the MUST rule, will DRS still kick in, and load balance resources within those 2 ESX host if one of the hosts is leading to 100% CPU or Memory utilization?

Yes, it will load balance the respective VMs within the host group constraint. E.g. if your DRS rule says "VM-Group-X" MUST run on "Host-Group-Y" (which consists of hosts A and B), then DRS will load balance the VMs in VM-Group-X between hosts A and B, but never on other hosts. VMs that are not part of VM-Group-X however will be able to move around in the whole cluster.

[Biga_b]

Ok. Got it..

This new SQL licensing model from MS makes the DRS/HA concept less flexible. licensing 2 esx host in one cluster is very expensive. We have 4 host in each cluster.

Summarizing:

With the MUST rule, i could create a active / passive situation within a DRS group if I disable DRS on the VM's in this group. Only in case of HA, the VM's will failover to the other host within the DRS group. Correct?

With the SHOULD rule, HA could place a VM outside the DRS group, and with DRS enabled, when DRS kicks in, this VM is going back into the DRS group.

Thanx

[MKguy]

With the MUST rule, i could create a active / passive situation within a DRS group if I disable DRS on the VM's in this group. Only in case of HA, the VM's will failover to the other host within the DRS group. Correct?

Yes, the approach with setting the VMs DRS level to disabled or manual would work. But it seems a bit odd not designating a 2nd host from the start. I mean how do you intend to handle host patching and or host maintenance? Do you really want to shutdown the VM in these cases? I think it's extremely likely that you will need the 2nd host anyways for some time, unless you are ok with some downtime.

I'm not too familiar with how SQL licensing works these days, but here are some nice guidelines if you haven't seen them already:

http://www.thelowercasew.com/new-sql-2012-licensing-and-its-impact-on-virtualization

http://www.vmguru.com/2012/08/microsoft-sql-2012-licensing-in-a-vmware-environment-part-one/

[Biga_b]

I just want to let my manager know all the possibilities. Also the wrong ones. And this was also a learning curve for me, to have a better understanding of DRS groups with the different rules in combination with HA/DRS. limitations etc..

I'm pretty sure we are going to license at least 2 hosts. Licensing every host would be better, but that's not going to happen. To expensive.

We have read a lot of articles the past 2 weeks about SQL licensing. We have to change our SQL licensing model from per CPU to per CORE. The extra cost gave my manager almost a heart attack.

Even with only two licensed host, we have to pay almost 80.000 euros extra. We have ESX host with 2 CPU's. 8 cores per CPU/socket

Anyway, thanks for helping, giving me a better understanding about the do's and don'ts surrounding Affinity rules