I like to draw attention to something that may not be well known and surprise some old ‘die hard’ DPM’ers. You probably are familiar with the ‘rule’ that all protected hosts within a cluster must be protected by the same DPM server! This is does not apply to Exchange 2010 and is formally documented here: http://technet.microsoft.com/en-us/library/ff399324.aspx. You still get the warning that all nodes must have a DPM agent installed but can be safely ignored for Exchange 2010. In that context you may want to take a peek at a related blog on resolving alerts; http://blogs.technet.com/b/dpm/archive/2010/06/04/automatically-resolving-alerts-such-as-resume-ownership.aspx
With other workloads including Exchange 2007 CCR the application store is considered to be shared across nodes in the cluster either physically or as replicated copy. DPM protection is not per definition fixed to a single physical node (although recommended for Exchange 2007 CCR). To maintain a single view on recovery points for a data source that from time to time could synchronize data using different nodes, these must belong to the same DPM server.
Exchange 2007 CCR uses different physical copies that are essentially ‘live’ synchronized seen by DPM as a single recovery source and therefore subject to the aforementioned rule. Exchange 2010 DAG’s are also synchronized copies but with a broader scope of usage such as ‘lagged’ copies. Hence DAG copies can no longer be seen as a single recovery source and it’s protection by a given DPM server now is a fixed pair. Nothing moves or switches across Exchange nodes in DPM context and therefore the aforementioned rule is no longer needed. This allows multiple DAG copies on different nodes within the cluster to be protected by different DPM servers. More importantly; this also allows protection of large Exchange configurations (>80TB of DAG stores) by multiple DPM servers.
Thank you for this analysis. We have a current environment running DPM 2007/Exchange 2007 CCR nodes and local DPM with remote SCR nodes. If I understand the analysis correctly, it is not possible to stand up a remote DPM secondary server (with agent deployed to the DPM Primary) and backup the Exchange mailbox database replicas to the DPM Secondary via the DPM Primary. Correct?
Following this logic chain, I take it that if I did stand up a DPM Secondary server facing the DPM Primary, the DPM Secondary host would not begin active backup of the mailbox database data and transaction logs unless I take the DPM Primary server down.