SRM and Exchange

[dcoz]

Hi Guys,

I am using a Lefthand SAN

I was curious about how people are using SRM to protected MS exchange.

• Is it better to use exchange's replication abilites like in exchange 2007 and 10?

• should you use RDMs to run VSS before replication?

• or can you place the exchannge server on a VMFS volume and normal replication be ok?

I guess these questions also apply to SQL as well.

Any help would be appreciated

DC

[admin]

Oh such a great topic...let's see...

•Is it better to use exchange's replication abilites like in exchange 2007 and 10?

HA vs. DR, it's that simple. CCR and DAG gives you highly available databases (in 2010) or highly available mailbox servers (2007). SRM gives you disaster recovery using your storage replication, block level replication, no database reseeds, no false-positive fail-overs, and can cover non-clusterable apps, non-MS apps/OS's, no Windows Server Enterprise license needed, etc.

•should you use RDMs to run VSS before replication?

In most cases SRM will use the last point in time replicated image to bring the services back online, so an application aware VSS snapshot, say 5 minutes before a disaster, might not be used because the last image, 2 minutes before the disaster, would be the default image used by SRM...the latest replicated point.

Now could you use RDMs to run VSS and use the replicated VSS based snapshot as an off-site backup...absolutely.

•or can you place the exchannge server on a VMFS volume and normal replication be ok?

Can you place Exchange on VMFS, yes. Will normal replication be OK? If by normal replication you mean storage based replication then, yeah. Exchange is extremely resilient, much more now that it has been in the past. I've seen many an Exchange server lose power, disk, or some other failure and recover as if nothing had ever happened.

In my 10 years of running Exchange for large enterprises I've experienced 2 database corruptions. If I were to run multiple databases, i.e. CCR or DAG, it wouldn't be because of a fear of database corruption it would be for high availability of the server and storage. Single-copy clusters are still popular for a reason.

-alex

[dcoz]

Thanks for feed back alex .

I usually setup exchange with OS on one VMFS, logs on another, and DB on third based on LUN RAID type.

In the underlying SAN replication i'm guessing the replication has be be identical, to allow exchange to be consistent on the DR side

How have you dealt with this?

Thanks again

DC

[admin]

Ahh..yes, you need to replicate at the very least the DB and the LOG volumes as one unit. Having the OS slightly off is ok, but I replicate the entire mailbox server as one unit. You don't want your DB to be at one point in time, expecting a log that the LOG volume doesn't have because the replication was too relaxed.

-alex

[dcoz]

alex,

so do you have the VM on one VMFS volume and not spread across various VMFS volumes?

Thanks for the help

DC

[admin · Accepted solution]

My VMs are spread across many volumes, average of about 14. All of those volumes replicate as a group to ensure consistency between the databases and logs more than anything else.

-alex

[dcoz]

Thanks for the help alex

Best Regards

DC