Adding Physical Disks Resources to an Existing Exchange Cluster   

I currently have a 4 node exchange cluster set up in an active passive 
configuration. Some of our message stores are becoming large and we want to 
add new storage groups to hold these message stores. I've read through lots 
of documentation on how to add more drives from a SAN but I still have a few 
questions. Any advice, kb articles, links would be a great help. 

1.) To add a physical disk resource do I need to shut of all other nodes in 
the cluster? Based on the document to have all nodes turned off, then use 
disk manager to initialize the disk and give it a drive letter. Then move it 
to other nodes to make sure they can write to it. I'm assuming this is to 
confirm that the SAN admins have put it in the correct group on their side. 
This isn't necessary for the health of exchange? 

2.) One of the disks I'm adding, I'm adding 2, will be not a drive letter 
but a mounted volume. Using http://support.microsoft.com/kb/280297 it only 
requires that you pause the cluster nodes then move it over to another node 
that will view it. Then you do In Disk Management, right-click the mounted 
volume, and then click Change Drive Letter and Paths. Is this on the root 
volume you are using or on the new drive you set up to mount as a folder? 

3.) When all Exchange Virtual Servers are brought offline, all the disks in 
disk manager show up the correct label on some backed end servers, but not 
others. Why is this? How do I make sure this has the correct label's across 
all back end servers. 

4.) Same thing for when its a mounted volume and it the actual Exchange 
service are brought offline, the disks just show up as basic disks that are 
formated with no label. How do I make sure they have the same label, I think 
question 2 should answer this. 

5.) When I open my computer when all disks are mounted, some servers see its 
active disks and the quorum disk and others only see its disks. The quorum is 
currently being held by the passive over node. Anyone have any ideas why 
this? 

Thanks

date: Tue, 11 Dec 2007 17:50:00 -0800   author:   MarkMorow

Re: Adding Physical Disks Resources to an Existing Exchange Cluster   

Hi,


> 1.) To add a physical disk resource do I need to shut of all other nodes 
> in
> the cluster? Based on the document to have all nodes turned off, then use
> disk manager to initialize the disk and give it a drive letter. Then move 
> it
> to other nodes to make sure they can write to it. I'm assuming this is to
> confirm that the SAN admins have put it in the correct group on their 
> side.
> This isn't necessary for the health of exchange?

You don't need to shut all the other nodes off in a cluster when presenting 
the disk to the first one to format it no. Don't present it to the other 
nodes till it's a resource in a cluster group though.

> 2.) One of the disks I'm adding, I'm adding 2, will be not a drive letter
> but a mounted volume. Using http://support.microsoft.com/kb/280297 it only
> requires that you pause the cluster nodes then move it over to another 
> node
> that will view it. Then you do In Disk Management, right-click the mounted
> volume, and then click Change Drive Letter and Paths. Is this on the root
> volume you are using or on the new drive you set up to mount as a folder?

I believe Step 5, where it says unpause the nodes, is just considered a best 
practice when adding new resources to your cluster. It stops any failovers 
for example happening, whilst you're working on the Active node.

> 3.) When all Exchange Virtual Servers are brought offline, all the disks 
> in
> disk manager show up the correct label on some backed end servers, but not
> others. Why is this? How do I make sure this has the correct label's 
> across
> all back end servers.

When the cluster is offline none of the disks should be accessible? What are 
you actually seeing? When the cluster is online whichever node holds that 
resource group will see the disks - and it should see the Volume Label you 
gave the disk.

> 4.) Same thing for when its a mounted volume and it the actual Exchange
> service are brought offline, the disks just show up as basic disks that 
> are
> formated with no label. How do I make sure they have the same label, I 
> think
> question 2 should answer this.

Again the disk are inaccesible, you shouldn't see anything about them.

> 5.) When I open my computer when all disks are mounted, some servers see 
> its
> active disks and the quorum disk and others only see its disks. The quorum 
> is
> currently being held by the passive over node. Anyone have any ideas why
> this?

The Quorum is within the default cluster group. Think of each 'group' as a 
seperate entity. Only the node that is active for that group will see the 
resources within it.

Oliver

date: Fri, 14 Dec 2007 00:29:22 -0000   author:   Oliver Moazzezi [MVP]

Re: Adding Physical Disks Resources to an Existing Exchange Cluste   

"Oliver Moazzezi [MVP]" wrote:

> Hi,
> 
> 
> > 1.) To add a physical disk resource do I need to shut of all other nodes 
> > in
> > the cluster? Based on the document to have all nodes turned off, then use
> > disk manager to initialize the disk and give it a drive letter. Then move 
> > it
> > to other nodes to make sure they can write to it. I'm assuming this is to
> > confirm that the SAN admins have put it in the correct group on their 
> > side.
> > This isn't necessary for the health of exchange?
> 
> You don't need to shut all the other nodes off in a cluster when presenting 
> the disk to the first one to format it no. Don't present it to the other 
> nodes till it's a resource in a cluster group though.

All of our LUNs are in the same group, I tried this in the test environment 
where I just re-do a disk scan on the one box, then add it to the cluster. It 
seems to work fine but will this cause any long term problems?

> > 2.) One of the disks I'm adding, I'm adding 2, will be not a drive letter
> > but a mounted volume. Using http://support.microsoft.com/kb/280297 it only
> > requires that you pause the cluster nodes then move it over to another 
> > node
> > that will view it. Then you do In Disk Management, right-click the mounted
> > volume, and then click Change Drive Letter and Paths. Is this on the root
> > volume you are using or on the new drive you set up to mount as a folder?
> 
> I believe Step 5, where it says unpause the nodes, is just considered a best 
> practice when adding new resources to your cluster. It stops any failovers 
> for example happening, whilst you're working on the Active node.
> 
> > 3.) When all Exchange Virtual Servers are brought offline, all the disks 
> > in
> > disk manager show up the correct label on some backed end servers, but not
> > others. Why is this? How do I make sure this has the correct label's 
> > across
> > all back end servers.
> 
> When the cluster is offline none of the disks should be accessible? What are 
> you actually seeing? When the cluster is online whichever node holds that 
> resource group will see the disks - and it should see the Volume Label you 
> gave the disk.

They disks are unaccessible but I think the idea was if your cluster is in 
that shape, for troubleshooting purposes you would know which disk is what. 
I'm going to take the poor mans way and just write out now all the disk ids 
and if my cluster is in that shape I'll probably be on the phone with 
Microsoft support and will at least have some idea which disk is what. 

> > 4.) Same thing for when its a mounted volume and it the actual Exchange
> > service are brought offline, the disks just show up as basic disks that 
> > are
> > formated with no label. How do I make sure they have the same label, I 
> > think
> > question 2 should answer this.
> 
> Again the disk are inaccesible, you shouldn't see anything about them.
> 
> > 5.) When I open my computer when all disks are mounted, some servers see 
> > its
> > active disks and the quorum disk and others only see its disks. The quorum 
> > is
> > currently being held by the passive over node. Anyone have any ideas why
> > this?
> 
> The Quorum is within the default cluster group. Think of each 'group' as a 
> seperate entity. Only the node that is active for that group will see the 
> resources within it.

Well for example if I reboot the box that usually holds the quorum, I move 
the cluster group to another node first. Reboot the box then move it back. In 
My computer on the box that held the cluster group for a few minutes, it will 
see all its disks and the quorum even after the cluster group has been moved 
back to the other node. I would never access the quorum disk that way but it 
is just weird that it shows up in that way. 

> Oliver 
> 
> 
>

date: Thu, 13 Dec 2007 17:25:01 -0800   author:   MarkMorow

Re: Adding Physical Disks Resources to an Existing Exchange Cluste   

Hi,

Inline,

> All of our LUNs are in the same group, I tried this in the test 
> environment
> where I just re-do a disk scan on the one box, then add it to the cluster. 
> It
> seems to work fine but will this cause any long term problems?

As long as you don't initialise the disk and start using it on other nodes, 
no.


> They disks are unaccessible but I think the idea was if your cluster is in
> that shape, for troubleshooting purposes you would know which disk is 
> what.
> I'm going to take the poor mans way and just write out now all the disk 
> ids
> and if my cluster is in that shape I'll probably be on the phone with
> Microsoft support and will at least have some idea which disk is what.

The disks will not be accessible, if worried, document their disk 
signatures.

> Well for example if I reboot the box that usually holds the quorum, I move
> the cluster group to another node first. Reboot the box then move it back. 
> In
> My computer on the box that held the cluster group for a few minutes, it 
> will
> see all its disks and the quorum even after the cluster group has been 
> moved
> back to the other node. I would never access the quorum disk that way but 
> it
> is just weird that it shows up in that way.

That is very strange, the only time i've ever seen this was on a vm cluster, 
there was obvious lag between Cluster Administrator on both nodes.

The disk should not be showing as accessible in My Computer however.


Oliver

date: Fri, 14 Dec 2007 12:01:48 -0000   author:   Oliver Moazzezi [MVP]