SRM 6.1 vs Stretched Cluster (vMSC)

[JonBelanger2011]

Hello,

A while ago were decided to use SRM for our disaster recovery tool for about 25 vms.  We use a mix of vSphere Replication and Array base replication (Currently using Asyncronous IP replication with a storwize v5000 and an older v7000).  It's been working well for us but the topic of stretched cluster keeps coming back in our internal discussions mostly to be able to balance workload from our main site to our recovery site.  After upgrading everything to version 6, I am now able to move VMs from one vCenter to another but I have to move storage also...  It works well, but the it would be great if we could simply use a standard vMotion and use the replicated storage instead.

We've recently acquired 2 new v7000 with hyperswap which offers syncronous replication.  I was told by IBM that in order to move workloads between our 2 sites, I have to eliminate one vCenter thus get rid of SRM.  This redbook also suggest a one vCenter topology.

From what I read in this SRM Whitepaper for version 6.1, I can have best of both worlds by configuring SRM with stretched storage.

I was wondering if some of you have configured SRM with stretched storage.  I'm curious to know how I could present the same volume to 2 vCenters.

Thank you!

[Cato Grace]

Stretched storage integrations with SRM is different from vMSC (what IBM is referencing when they mention transitioning to a single vCenter).

Stretched storage integration with SRM requires specific storage from your array vendor that supports not just synchronous replication, but also the ability to present the same LUN/datastore to both sites/arrays as read-write at the same time AND an SRA that will work with that array and SRM 6.1 or later.

For details on stretched storage in the SRM documentation see here: Site Recovery Manager 6.1 Documentation Center

For compatibility information see: VMware Compatibility Guide - sra

Note that there is a filter option for stretched storage. When I search for IBM and SRM 6.1 w/Stretched storage I see the Storwize SRA ver 3.2 and 3.2.1.

[JonBelanger2011]

Hey gs_khalsa,

Thanks for your reply.

SRA ver 3.2.1 for Storwize is already what we are using for ABR with SRM 6.1.

How do I know if stretched storage integration is also supported with an hyperswap volume.  When looking at the search results in the compatibility guide, I only see Array Model(s) : SVC

Thanks again for your help!

[Cato Grace]

That would be a question for IBM. The SRA documentation might help?

[JonBelanger2011]

gs_khalsa

Do you know of an array vendor that supports stretched storage integration with SRM 6.1?  So far, all the information I find about Hyperswap points me to a vMSC.

From IBM : "There is no SRM type plugin for HyperSwap at all since, HyperSwap circumvents the need of such a plugin. It is the solution to get rid of the expensive solution with VMware Site Recovery Manager. Same is valid for any other supported host type with HyperSwap, like Windows with Hyper-V."

Thanks!

[Cato Grace]

Kind of strange that IBM would say that after being one of the vendors that has an SRA that supports this and has publicized it: Considering "VMware SRM stretch" ?? – Think IBM San Volume Controller (SVC) - Storage ISV Blog - Storage ISV

It sounds like they are pushing vMSC over SRM integration so I would recommend you look into the differences.

For HyperSwap have them explain to you how would:

- Non-disruptively test a failover

- Orchestrate a disaster avoidance scenario

- Run scripts or have callouts as part of a recovery plan

[JonBelanger2011]

gs_khalsa

I guess it comes down to Hyperswap vs SVC ..

From what I understand, SVC is a more robust and expensive solution.

[Cato Grace]

I'm not really familiar with either. If hyperswap is what you have it sounds like what you'll need to go with unless you can justify moving to SVC. You can use my list of items as justification if that helps.

[reaverza]

You say "it would be great if we could simply use a standard vMotion and use the replicated storage instead" and "best of both worlds". I am curious why you think that it's better to have two vCenters with stretched storage?

One massive advantage of a single VC, is being able to cluster your host together across sites and use HA, which effectively gives you an active/active DC configuration. In such a scenario I don't even know why you would bother with SRM.

[JonBelanger2011]

We already have SRM and it's been working well for us.  We use the SRM Test feature quite a bit.  A stretched cluster doesn't give us the ability to test a failover in the way that we need it to.

Being able to use cross-vCenter vMotion with stretched storage and have SRM at the same time for orchestrated DR is what we thought we could achieve with Hyperswap.  From what I understand, SVC can do this but not Hyperswap.

I'm not against a stretched cluster down the road.  But for now, SRM is in place and we don't have all the compute power at the remote site for a stretched cluster.  Stretched cluster also means double the storage which we don't have at the moment.

Like @gs_khalsa says, a stretched cluster doesn't give me :

- Non-disruptively test a failover

- Orchestrate a disaster avoidance scenario

- Run scripts or have callouts as part of a recovery plan