2011-11-17 03:30 AM - edited 2015-12-18 01:23 AM
I am looking to migrate from exchange 2003 to exchange 2010.
I have created 3 exchange 2010 server
I plan on creating a DAG, but I am not sure were snapmanager is installed? would I have to install it on both servers holding the mailboxes? could I create another server running the mailbox role and have that hold all the passive databases for the DAG, and install snapmanager that?
2011-11-17 02:19 PM
SME has to be installed in all mailbox servers regardless if they are passive or active nodes. You then have the option to either backup the active or passive or both nodes at the same time or at different schedules. If you back up the DAG and not each individual mailbox server you also have the option to backup let's say the passive node and perform a copy based backup of the active node and then you can perform the verification process at a different interval should you choose to do so. Look for the SnapManager 6.0.2R1 for Microsoft Exchange Release Notes document released on May 2011.
2011-11-18 02:34 AM
Slightly confused by your reply cserpadss. You start by saying that SME has to be installed on ALL mailbox servers regardless then state you can backup the DAG rather than the individual mailbox servers.
If this is the case then surely you can get away with just putting SME on a DR DAG member an snap / verify the DAG down at DR?
2011-11-18 03:01 AM
This is what SME6 Installation and Administration Guide says:
Prerequisites for migrating Exchange Server 2010 mailbox databases
[...]You have to install SnapManager on all member servers of the Database Availability Group (DAG) before migrating the mailbox databases. If SnapManager is not installed on a member server of the DAG, all databases hosted by that member server, including both the active and passive databases, will not be migrated by SnapManager, and SnapManager will not be able to back up databases on that server.
2011-11-18 06:13 AM
In agreement with some of the other posts, we've installed SME on all mailbox servers on both the active and passive side of the DAG. The "functional" side of this is that if we were to sustain a long"ish" outage in either our passive or active side and were running exclusively on one set for awhile, then we've got SME jobs already set up and won't need to worry about hustling to get the schedule set up in a crunch. Additionally, we're usually operationally performing SME restores on our primary site (when needed) but we're then leveraging the SME backups on our passive site to perform NDMP and/or SnapVault backups (thereby taking the read "hit" on our passive controllers). Having both sides set up with SME lines you up with the install/admin guide and gives you some flexibility and protection in your overall Exchange operation.
2011-11-18 08:15 AM
It depends how you want to perform backups. Do you want to backup all the members of the DAG or an individual mailbox server? For example I have 3 members in my DAG, 1 in my DR and 2 in my active site. When I do a backup of my databases I do it in all 3 of them, at the same time and I do the verification later in the day at the DR mailbox server.
If you only want to backup 1 mailbox server then you just have to install SME in that server. The other members of the DAG will not be backed up which may be fine you just won't be able to use the gapless DAG backup copy functionality and also you will not be truncating your transacation logs at the mailbox server not running SME. Are you using cicular logging if not how do you plan on truncating the log files on the members of the DAG that aren't getting backed up?
Are you going to be using other NetApp functionalities such as SnapVault / SnapMirror between your active/passive sites?
2012-01-18 02:23 AM
I have a dilemna with what to do with the truncating of logs. I have a 4 node DAG (two in prod site / 2 in DR) and don't use snapmirror. I want to truncate the logs but don't want to run the verification process on the live DB vols for performance reasons. The delayed verification process doesn't appear to flush the logs so I'm stuck with maybe having to run the veridfication backup out of hours. The only issues with this is that if the verification process fails I could end up with a log build up the following day which could end up dismounting the stores...