Exchange 2007 CCR convert into ESX 4U1

[b4ndit]

dear all,

I'll convert a Exchange CCR cluster in the next month.

It has 2 Hub+CAS server with 2 Mailbox server. HUB&CAS server are load balanced and Mailbox servers are clustered.

Some articles says that is not recommend to convert Exchange Mailbox server into Virtual World because its high load (performance may degraded issue).

It's true ?

And, I would like to know some other consideration must be taken before this Exchange 2007 CCR converted to Virtual Wolrd..

note:

HUB&CAS using Win 2008 STD 64x (utilization : 5-7% CPU,70-75% Memory)

Mailbox using Win 2008 ENT 64x

Many thanks for your responds

BR,

Gari

[Dr_Virt]

First, let me reassure you that running Exchange 2007 with CCR in vSphere 4 is not an issue if properly deployed. I currently do so and have assisted clients in deploying the same.

But, your post does not give many details. A thorough analysis must be undertaken to ensure sufficient performance and reliability. Some questions that must be answered are below:

1) How many mailboxes and associate messages

2) Daily I/O pattern (peaks, gross mesage volume, etc.)

3) Storage design

4) Future growth

5) vSphere cluster topology (is there one, how many hosts)

6) vSphere host resources (CPU, memory, etc.)

7) Other guests on planned hosts. Are there any other guest VMs that could create contention for resources?

😎 Perfmon - get to know your physical systems performance and needs very well.

Remember, when virtualizing, you are sharing slices of the pie in a dynamic way. My Exchange 2007 systems consume all resources provisioned to them. Expect minimal sharing when the Exchange VMs are active. If 8GB of RAM is provisioned, they will use 8GB of physical RAM.

To sum up; plan, plan, plan. If you have planned well, it will run very well.

[b4ndit]

thank you dr virt.

Hearing your answers, it's reliefs me

I'll try to check & provide these measurement.

Thanks

[murphyslaw1978b]

I wanted to jump on this thread, even though it's an older one.  I have a 2-node Exchange 2007 mailbox CCR setup on ESXi 4.1 with full VMFS datastores.  We have about 3,500 mailboxes, 8vCPUs and 32GB of RAM on each VM.  Everything is running and working fine, but when I need to do maintenance on a host, I normally pause the passive cluster node, migrate it using vmotion, and then resume.  The last time I did this, the active node went down.  I know that v-motion is "not recommended," and I can now see why VMware would say this, but my question is why this happend.  I thought pausing a microsoft cluster node ensured that the node would not try to take over.  My other question is how do I migrate the VM off?  Should I shutdown the passive node and then do a cold migration to another host?

[AlbertWT]

Hi Murphy,

About the datastore, are they using normal VMDK disk or RDM with Virtual compatibility mode ?

What happened if you just shutdown the passive host rather than pause it (PAUSE as in VMware level from vSphere console) ?

[murphyslaw1978b]

The datastores are all VMFS - no RDMs.  However, I was not clear when I said "paused."  When I pause the Exchange passive node, I'm actually not pausing the VM in VMware.  Instead, I'm going into the application, launching Microsoft Failover Cluster Management, and selecting the passive node, right-clicking it, and pausing from there.  This is the recommended way that Microsoft wants when performing maintenance on an MS Cluster.  I've learned, however, that pausing a Exchange cluster node neither stops CCR replication, nor does it fail any resources.  So pausing a node appears to not be the end-all, be-all way to perform maintenance...

I am planning on pausing the passive node, and then shutting down the passive VM.  I need to move the VM to a different DRS cluster node located in a different data center, and I also need to migrate the storage to a new SAN.  So once that's done, I can startup the passive node and resume it so that exchange database replication resumes.

[AlbertWT]

Murphy,

AFAIK you need to move the cluster mailbox server from Exchange management console and then stop the Storage Group replication as per this article: http://www.msexchange.org/articles_tutorials/exchange-server-2007/high-availability-recovery/deployi... rather than using the failover cluster wizard at the MSCS level.

[murphyslaw1978b]

Yes, I had thought about that as well.  But in most of the Microsoft documentation regarding performing maintenance on Exchange 2007 clustered mailbox servers, I see only that Microsoft recommends to perform maintenance on the passive node.  I don't see anything hard and fast that says that replication should be stopped.  Although it makes perfect sense to do that of course.

[murphyslaw1978b]

I stopped Exchange replication, paused the cluster (in MSCS), and then shutdown the passive node for a cold migration.  This worked and caused no impact to the active node.  When it was powered back on and replication resumed, the databases were healthy and replicated within minutes.  I will be performing these cold migrations going forward, until ESXi 5 and Exchange 2010 SP1 are out.  These two combinations - I believe - are supported for full vmotion.

[AlbertWT]

Yes, that's is also what I've been doing with my ESXi 4.1 Exchange 2007 SP1 CCR.

Glad that everything seems to be working fine with you too.