Does VMware officially support HA across WAN?

[XaTechAruba]

Hi All,

I am trying to find official documentation whether VMware either supports or does not support HA across a WAN. I have already searched but could not find anything. Could anyone could point me in the right direction?

Thank you,

Elbert

[plynch01]

XaTechAruba

Perhaps you could some further information as to what you are trying to achieve and with what version of vSphere, for the community to provide some further assistance.

Assuming that you are looking to provide a failover capability across a wide area network through utilising vSphere High Availability - for example, one of the main requirements for HA to function correctly is visible, shared storage between all hosts in the cluster. So, to achieve such a process does potentially offer a number of potential issues from the onset.

As I understand it, VMware HA does not support across WAN migrations, as for technologies like vMotion and HA to correctly function - nodes should reside in the same LAN segment. However products such as vSphere Replication and Site Recovery Manager offer an enterprise solution to provide cross-site fault tolerance.

I hope that helps in the first instance.

[XaTechAruba]

Hi,

I am actually verifying best design practices regarding DR Site setup using VMware vSphere 5. We are a IT consulting firm based in Aruba, and are in the process of submitting a design/proposal for a potential client. It seems that another consulting firm has recommended the client to use what VMware calls a "Stretched Cluster" design, extending VMware HA into the DR site via a high-speed fiber link. Now because they are only relying on one SAN at the production site in that design, if the production site goes down (not only because of a failure, but could also be natural disaster or any other scenario) they will not have access to their storage, defeating the purpose of having a DR site.

I have read some documentation over stretched cluster vs SRM (which is what we are recommending), such as http://www.vmware.com/files/pdf/techpaper/Stretched_Clusters_and_VMware_vCenter_Site_Recovery_Manage_USLTR_Regalix.pdf

but was just wondering if VMware actually supports HA being used that way.

Kind regards,

Elbert

[plynch01]

Hi Elbert,

Thanks for the clarification, that's a lot cleared now what you are trying to achieve - certainly sounds an interesting design.

My understanding is that specific stretched or metro clusters are supported with vSphere 5, but there are a number of specific recommendations that need to be followed in terms of architecture and there are some limitations with regards to capability without the use of any form of storage replication between sites.

In the scenario you mentioned, am I right in assuming that you are intending to only deploy the vSphere hosts in geographically diverse locations with a single storage array based in the production site. All storage I/O will be served from the production site, even to the secondary site over the inter-site connectivity (so effectively a L2 extension)? Is that correct?

Kind regards,

P.

[XaTechAruba]

Hi Pierce,

You are correct, this scenario has been recommended by a 3rd party, using 2 diverse locations connected via layer2, and with a single storage array in the production site. We have recommended to implement a DR site using VMware SRM, using vSphere Replication to either onboard storage on the servers, or a secondary SAN using storage replication, although the client does want to keep the DR site as cost-effective as possible.

Elbert

[plynch01]

Morning Elbert,

To be totally honest, I am not sure if a single storage site topology with multi-site HA is supported officially. I personally have never designed and deployed an environment based on the scenario you have described - only ever a true multi-site topology with dual storage arrays and replication between etc.

Generally, I can't see why this wouldn't be a supported architecture as long as it follows the basic requirements with regards to the inter-site L2 connectivity (ie RTT less than 10ms to support HA/vMotion) but this is something I guess VMware will have to confirm.

I've had a bit of a wild thought though (warning - this will require further validation :smileycool:), if you aim is to essentially provide a high level of fault tolerance with an active/active multi-site topology while attempting to reduce cost - you could investigate something like replicating from your primary site to a vSphere Storage Appliance setup? My thinking here is that, your scenario will potentially work fine until there is a major failure at your primary site (be it the single storage array scenario being ultimately a single point of failure) - so, what this might allow you to do is provide an extended level of faul tolerance in the event of the primary site failing so that you have some form of active storage available in the secondary site.  Just a wild suggestion, and will require a lot more thought and consideration!

Hoperfully some other community members will pick up this thread and offer their thoughts and advice with regards to the officially supported multi-site architectures... :smileyhappy:

Regards,

P.