Is SRM needed?

[zjan79]

my architecture is like the pic above. all esx are considered in 1 network (production) with 1 SAN storage. the SAN storage replicates to the other on the other site.
my question is, if the left side network (esx1, esx2, Storage, VC) goes down. how is my vm's still able to function properly.
I know that you dont need VC to work in order for your VM's to be booted in ESX3 or ESX4, how about the storage? with just both SAN's replicated, can it still work ??

[Linjo]

What are your requirements and business needs?

// Linjo

[idle-jam]

yes with both replicated you can mount the VMDK and power up. however it will be very manual and one by one VM to do. in the end you might not mean your business objectives.

with SRM you will need vcenter at both sites. also you will need to configure san replication.

[amvmware]

You first need to understand the business requirements  - this will then drive the technology requirements. For a DR strategy you do not need to deploy SRM, but as has been indicated the manual processes could be time intensive depending on the scale of your environment.

You need to understand the key business applications and the impact to the business of these applications not being available.

I would also be looking to build in resilience for your VC server as part of a DR strategy. SRM requires a second VC.

[zjan79]

thanks for the feedback.

i roughly got the answer from a vm partner, but just wanted to get your advice.

[bulletprooffool]

Hi,

you have a few options, depedent on budget and time /  skills, as well as the DR scenarios you'd like to coover (e.g., if you'd  like to accomodate full site failure, you will need to have something  in place that replicates all pieces of the puzzle)

Additional  considerations are things like Network (Do you have a stretched VLAN,  or will IPs need to change in event of failover) and so on.

1)  The easy (but expensive) solution is SRM

2) In the event of site  A crashing, you can manually 'flip' your storgae to site B, then import  all vmdks to the B Side cluster. You'll need to make sure that Port  Groups, etc are all aligned with Site A

3) You could script the  failover (This is pretty easy if you have someone on site with the  correct skills and pretty cost effective)

Usually, the  biggets challenge is the IP addressing.

There are  several variations, which will require use of further VM technologies  (FT etc) - what you need is a consultant in for a day , that is  independent of any sales etc . . that can look at your full solution,  all your requirements, your budget and come up with not just one easy  and expensive solution, but some alternatives aligned to your specific  environment.

If your VM Partner has not asked the following questions, then I am guessing they have only offered SRM as a tool:

Do you have stretched VLANs?

What is your budget?

Do you have scripting skills in-house?

What is your Recovery time objective?

What is your recovery point objective?

How Many VMs / Hosts will need to failover?

What is the link speed between sites?

What amount of data are you replicating?

what technology are you using for replicating data?

Are all Datastores replicated?

Do you need a solution only for full site failover, or for maintenanace failovers tec too? (e.g. Migrate Site to site for planned works etc)

Is ESX hardware at all site of same EVC compliance?

Unfortunately, VM Partners make no money from telling you to script a solution . . so the list of options they proivide can be very limited.