SRM Failover Scenarios

[gboskin]

We are looking at implementing SRM . However we have two failover scenarios

Senario 1

Complete datacenter failover to DR site (here am sure SRM will work fine)

Senario2

Partial failover. Here we would like to failover a couple of virtual machine for what ever reason to ur DR site and still live other virtual machine on the production site. We will thus have images running on both site . can SRM do this ?? can it fail back the images back to the production site without disrupting the running images ??(Sounds messy)

[jbruelasdgo]

1- no problem

2 - it really sound messy because you are telling that you want to failover and keep running both source and target (it sounds like you would need to have DNS, AD, services, etc duplicated which I think could be dangerous). you definitely can failover and then create a failback scenario, but keep it running both source and target??? (it really sounds messy) (you would need to take the VMs out of the protection groups so they can "run", I mean that they are not running on replicated disks that are actually being used by the replication solution, when SRM is running it depends on replication at storage hardware level, so that would be a question for storage admins, can the replication allow to access both the source disk and replicated disk at the same time??)

regards

Jose Ruelas

[gboskin]

yes and No.

what I mean is that lets say we have 100 VMs and only 10 have a problem and we want to failover only just the faulty 10. and not the entire datacenter..

Can we just failover this 10 and fail them back when the issue has been resolved. Do the other 90 still run on the production site replicating as usual??

I guess we will have to have the 10 in the protection group??? (Initially all 100 should be in the protection group and then if we need to do a partial failover take out all the good running VMs out of the protection group)

The SRM replication adaptor can it handle this?? I guess it should since replication can be done on the LUN level

Am just thinking aloud... going to give SRM real work over test soon

[jamieorth · Accepted solution]

Depending on your storage provider, how you have your replication setup (I think you said individual luns??) then SRM allows you to create as many "Plans" as you want. So in theory you could have a plan that fails over a particular set of VM's as long as your storage would allow you to fail just those over without impacting the other 90 that you don't want to. Does that make sense? If you really boil it down, all SRM is doing is allowing you to automate some of the steps of failing over VM's - if you can do those 10 manually without impacting production, then SRM will do it as well. DESIGN is the critical component. We are finding out that we are going to totally redo how we have our LUN's and replication configured to give us the most flexibility when it comes to SRM.

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...

[gboskin]

Yes thats bravo ..jhust waht i wanted to hear ..yes design is surly the key ..we look at creating 3 level of protection gropus/plans

one for invidual VMs , one foe each resource pool and one for each cluster

Anyway we will be giving SRM a real test in the next coming weeks/months before we go live

[jamieorth]

We just got ours up and running (in the lab) yesterday and today. We had to wait on EMC to release an updated SRA for our DMX. We took a VM on our test lab lun and we have flipped it over several times today, fun stuff....

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...