fonzie2k
Contributor
Contributor

VMware design - SRM, vMotion and redundancy

Greetings VMware geeks,

I am currently consulting a company with VMware, and i've met some .. uncertaints on my way, and i would like to hear if the community has any advice Smiley Happy

The current setup is as following:
6 hosts, 2 clusters, 2 NetApp SAN's and 2 vCenters.

Configured as 3 nodes in each cluster, and each cluster in its own datacenter on its own vCenter server.

The vCenters does not use linked mode. So there is no communication between the two sites at all.

The current logical setup is described in this picture: http://torgeirarnoy.no/Current_vmware.png

Now, with this setup its not possible to use vMotion and its no redundancy at all between the two datacenters.

The physical setup is two datasenters, but they both have the exact same network setup, and a very high network capacity between eachothers.

(So there is no bottlenecks in sight here)

So i dont see why there should be two vCenters in  the first place, right? Why not setup the two clusters in the same vCenter server using the same datacenter (vCenter datasenter)?

We (they) also got a Site Recovery Manager (SRM) license for 25 VMs. That could be used for disaster recovery between the two physical sites, but still, no vMotion.

So my idea is:

* Cross mount the NetApp disks so all 6 hosts can use them (required for vMotion).
* Use only one vCenter and attach the 2 clusters under one datacenter (vCenter datasenter).

* Make a datastore cluster out of theese two NetApp disks to replicate data for redundancy.

Here is an drawing on how i might see the optimal setup: http://torgeirarnoy.no/Vision_vmware.png

The problem:
* No redundancy on the vCenter server. What if the site that holds the vCenter server goes down. Will HA etc continue to work?

Any advices here will be deeply valued.

Regards, Torgeir.

0 Kudos
9 Replies
BostonTechGuy
Enthusiast
Enthusiast

If you are using SRM, then a disaster recovery discussion is happening?  If so the you should not do your second design.  SRM is the game changer here in this discussion.

The design that has each datacenter with its own infrastructure is the way to go for SRM.  Also if you are going with 5.1 or planning to go with 5.1, then you should place a SSO and Web Server at each Datacenter.  You only need one per physical site.

Linked mode also helps to make it easier to manage the vCenters.  Single Pane of Glass idea.

Check out the post that I replied to yesterday for someone else on SRM accounts, but the design would be the same for you.

http://communities.vmware.com/message/2232522#2232522

Also I have a question for you:  Where did you get those VM shapes?  They are fantastic.  I would like to use them in Visio for my clients.

Thanks,

BostonTechGuy.

mcowger
Immortal
Immortal

Your problem is here:

"* Make a datastore cluster out of theese two NetApp disks to replicate data for redundancy."

Datastore clusters do not replicate data.

--Matt VCDX #52 blog.cowger.us
fonzie2k
Contributor
Contributor

Oh, ok. I thought there was some settings making a datastore replicate data accross all its disks.

So, my only alternative to make theese two clusters redundant, is to have two vCenters and use SRM?
And that means, i cant both have vMotion between clusters and redundancy?

0 Kudos
fonzie2k
Contributor
Contributor

Here you can find the Visio stencils i used, BostomTechGuy: http://communities.vmware.com/docs/DOC-11498

Again, thanks for all feedback so far.

0 Kudos
mcowger
Immortal
Immortal

Oh its possbile to do stretched clusters with NetApp, its just FAR more complex than you are thinking it is.  It involves lotsof extra disk, putting one of the 2 controllers on each side of the link (1 controller per site), etc.

Talk to your netapp reseller.

--Matt VCDX #52 blog.cowger.us
0 Kudos
BostonTechGuy
Enthusiast
Enthusiast

Plus if you are having a Diaster Recovery plan discussion, using the same NetApp Filer would defeat the purpose.  Its physically putting all your eggs in one basket.

Thanks

BostonTechGuy

0 Kudos
IsmaelRossell
Contributor
Contributor

Hello all,

I have a similar aproach.

The current setup is as following:

2 Datacenters in the same campus, 4 hosts (2 on each datacenter) , 2 Netapp SAN's ( 1 on each Datacenter), 1 vCenter, 1 Cluster with 4 hosts.

Both Netapp's are accesible by 4 hosts ( 8Gb FC) this allow vmotion and storage vmotion. LUN's on Netapp-1 are replicated on Netapp-2, and viceversa.

The goal is power off datacenter-1 and give service with datacenter-2 without registering all the vm's again.

My problem is, after I have presented the replicated LUNS to the alive hosts, I cannot import the datastore without re-signaturing because a datastore with the same name exist.

Any advices here will be deeply valued.


Best regards,

Ismael.

0 Kudos
eduardserkowits
Contributor
Contributor

i have no access to this article...
can someone provide it here?

 

Cheers

0 Kudos
scott28tt
VMware Employee
VMware Employee

@eduardserkowits 

Which article?


-------------------------------------------------------------------------------------------------------------------------------------------------------------
VMware Training & Certification blog
0 Kudos