Stupid question here.. Why don’t use let Microsoft do the replication?  I used that in my former gig (e2k07, on 2 different heads FC, 7-mode).  Failover was (hate to say it) magical.  We also had the logs on the same location as the backups as we generated over 100GB of logs a day, so hard links can be your friend.
  Maybe its a MS cluster thing you are not doing, and that’s why you can’t do it.  I am no exchange expert, but the exchange folks had to make sure replication was always working, and we used SME for the backup/snapshots and had Netbackup mount the snaps for the secondary exchange server.


Thanks


Steve Klise


From: <toasters-bounces@teaparty.net> on behalf of Milazzo Giacomo
Date: Wednesday, August 5, 2015 at 9:34 AM
To: "toasters@teaparty.net"
Subject: Virtualized Exchange 2010 on cDOT

Hi all,

 

we’re going to migrate a running environment in 7Mode to cDOT and, nevertheless I’ve studied all docs about best practice and DR scenarios I’ve a lot of doubts about best architecture.

 

Now we have.

 

Production site.

2 E2K10 virtualized Server with a two copies of db in DAG environment

These 2 servers have their data on different controllers of an HA system and db are balanced on two controllers/server some as active and some as passive (but you can imagine also one server running active copies and the second the passive ones)

Servers are this way built:

C:\ system disk with Windows and Exchange binaries on a NFS vmdk with vmx in the same path

Database disks on VMware iSCSI raw devices (this because Microsoft does not support NFS vmdk for its data…no comment!..we know…)

For obvious reason there’s also another little iSCSI datastore containing the “pointer” to those raw device, the mapper (rdm). This is the way it works

Raw devices are managed by SnapDrive for there’s also SME running for backup/mirror.

 

DR site

At recover site there’s an identical enviroment with all the volumes replicated using SnapMirror modulated by SME

VMware SRM together with SRA mount, in case of disaster, the right vmdk, LUNs and so on using its own script so the E2K10 servers at DR can start with all DAG

 

One of the constrain of SRM/SRA in 7Mode is that ALL the volumes with NFS vmdk, iSCSI raw devices and iSCSI ds with rdm MUST belong to the same controller; so that is not possible to spread volumes on different controller/aggregates maybe to increase resiliency…Otherwise SRA will fail.

 

What’s about cDOT?

Can I create two different SVM, one just for NFS vmdk for binaries and the second one for iSCSI contents? Or have/may have I to put all togheter?

Have I to create, as stated in DR scenarios TR from NetApp, two SRM to manage the two Exchange servers or, maybe, to relocate data on different aggregates? Have these two aggregates to be resident on the same node or they can belong to the same node?

What’s about SRM/SRA constrain?

 

Thanks a lot for any suggestion/idea.

 

Regards,.