Exchange Cluster

[rondebbs]

Hello, I'm familiar with our SQL Clusters but new to our Exchange clusters. On SQL we have three node clusters that run a total of four instances in an active, active, passive environment. The two active nodes run two instances of SQL. If either active box fails it can fail over to one of the other nodes. If two boxes fail then all four of the instances can run on the existing node. This gives us lots of flexibility with SQL.

For Email we run Exchange Server 2003 Enterprise on Windows 2003. We also have three node active, active, passive clusters (node a, b, c) but with very little flexibility. Node a and b run one instance of Exchange while c is passive. If a or b fail they can only fail over to the passive node c. Node c can only handle one instance so if a and b both fail one of them will have an outage. If node c is running an active instance then we have no failover because nodes can only fail to c if it is not running an instance. Nodes can never fail to a or b.

This inflexibility seems crazy to me. The Excange guys say that is the way Exchange clustering works. Is this true or is it possible to set up the Exchange cluster similar to our SQL cluster where any node can failover to any box in the cluster.

 

[WhoKilledKenny]

Your Exchange Admin is correct with the setup. An Exchange server is limited to 4 storage groups per server(not counting the recovery storage group).
Your config may be as follows:
EVS1/ClusterNode1 (two storage groups) Active Node.
EVS2/ClusterNode2 (two storage groups) Active Node.
ClusterNode3 (no strage groups) Passive Node.
In this setup, both EVS1 and EVS2 are able to fail over to Cluster Node3. 2 SGs + 2SGs = 4. All 4 Storage groups can run on one node.

Another possible config:
EVS1 - 4 Storage Groups
EVS2 - 4 Storage Groups
ClusterNode3 - Passive
In this example only one Exchange Virtual Server (EVS) can fail-over to the passive node. If the second EVS attempted to fail-over at the same time, 4+4=8 storage groups, it would fail as it would be greater that the 4 SG limitation.

Both solutions are viable options.

 

[rondebbs]

Yes, your right we have 2 storage groups on each of the two active nodes. Is it possible to a forth node to this cluster? This would allow both active nodes to failover in the event of a double failure.

 

[WhoKilledKenny]

Yes, you could add another node to the cluster. But since you have 2 SGs on each virtual server (EVS1 and EVS2) both virtual servers would be able to fail over to the passive node, as the passive node can handle 4 storage groups total. So it sounds like you are already configured for that.

 

[rondebbs]

yeah, I'm not sure why our Exchange guys say that one node can not run both instances. Maybe they feel one node does not have the power to run both or maybe it's because each node has a recovery group also?

 

[WhoKilledKenny]

An Exchange server is limited to 4 storage groups per server(not counting the recovery storage group).

What I meant here was that the recovery SG doesn't go against the total.
So if you had 2 SGs on each node that would be 4 SGs and 2 Recovery groups = 6. But as far as clustering goes, it is still only 4 Storage Groups. So all 4 storage groups would be able to run on one node.