home links tools blog about
home

« It's alive | Main | Thank goodness for Virtualization »

August 01, 2005

Comments

Michael K. Campbell

UPDATE: (Updating this so that aggregators don't have to suck down that massive post just for a few changes) Section 4 points out that moving to your nodes to the domain may force the NICs to DHCP. Actually, looks like this actually a bug with VMWare - so keep checking your NICS periodically. Definitely check each node before you attempt to add it to the cluster. (And you'll probably see something goofy along the lines of windows saying: Hey, the addy you is specified is already in use on this box, by a secret NIC, do you want to stare at the ip addy screen (YES), or override and set this IP address (NO). You'll want to select NO.)

Daniel Melameth

…just sending big thanks for your concise and complete instructions on setting up a cluster in VMware 5—I found your bog “googling” for vmware and quorum and I’m glad I did. While the Cluster Administrator correct found my quorum drive, I also ran into the issue you did with the NICs having issues—for some reason, the private and public interfaces kind of switched and I had to reconfigure them. Quite annoying, but the resultant “pocket cluster” is well worth it.

Mark Johnson

I too want to thank you for providing such a clearly laid out process to perform SQL Server clustering in VMware. I only slightly modified your steps by creating the three VM servers in a Team (AD, Server1 & Server2) using a LAN segment (LAN 1) for the heartbeat network and NAT for the External. I also used a linked clone of a Windows 2003 base for each of the three servers. I too had the same issue involving the re-configuration of the external network's LAN interface. For me, it occurred after manually creating the .vmx file's disk entries (which I had to do with all servers shut down and VMware closed). It seemed to force a hardware re-discovery in Windows after starting everything back up. But once I changed the static IP settings back (and pressed "NO") it was smooth sailing from there. Therefore, I can confirm that it's still possible to use this method with VMware version 5.5.1 build-19175.

Michael K. Campbell

Ed Gummet (in the UK) sent me an email with what looks like a solution to the NIC Switching evil going on up in Step #4.
His solution:
Add the following line to your .vmx:
Ethernet0.virtualDev = "vmxnet"

Thanks Ed!

Rick

Does this still work with VMware Workstation 5.5? I started the VM and received a message that clustering was not supported and that the settings would be ignored.

Peter Ciaccio

Thanks for the guide but I've still got one remaining issue which is killing me. I can successfully create a shared disk which both servers can access (before creating a 2 node server cluster). The problem is that if I create a file on Server A it is not seen by Server B. The only way server B can see the new file (or change to a file) is to un-mount the disk in disk manager and re-mount it (like a refresh). I'm not sure if this is vmware issue or 2003 server issue. Is there something with the OS's that I need to do to make them aware that the disk is shared?

NB - I'm using 'Vmware Server Beta'

Any help will be greatly appreciated.

Pete :D

Michael K. Campbell

Peter,

It would appear that VMware Server Beta has modified something in the way disks are shared. I would suggest dropping a link to this post off in the forums, and ask to see if anyone has any ideas for how to fix the problem. (I've not had a chance to play with VMware Server Beta yet... though I might do so in a few weeks...)

(Sorry.)

Jay Wood

Just another "Thank you" for the well documented step-by-step procedure. We're using VMWare Workstation 5.5.1 build 19175 and your steps worked fine for me also.

-I used an existing Domain Controller in our lab instead of creating a new one with VMWare.
-Ed Gummet's tip to avoid the NIC issue in step 4 worked for me.
-I also get the message about clustering not being supported when I start either of the VMs, but it appears to have no affect. I assume it is just a warning and nothing else.

Thanks,
j

Michael K. Campbell

As an FYI:

I've just completed the 'clustering' process on VMware Server (beta).

A few things have changed in VMware Server... (like the removal of cloning functionality, and decreased snapshot functionality), but everything worked fine.

Well.. I did have to make a few tweaks to the disk configuration. Here's a sample configuration of what was needed:

diskLib.dataCacheMaxSize = "0"
disk.locking = "FALSE"

scsi1.present = "TRUE"
scsi1.virtualDev = "lsilogic"
scsi1.sharedBus = "virtual"

memsize = "756"
scsi0:0.present = "TRUE"
scsi0:0.fileName = "Windows Server 2003 Enterprise Edition-000001.vmdk"

scsi1:0.present = "TRUE"
scsi1:0.fileName = "F:\Quorum Disks\Quorum.vmdk"
scsi1:1.present = "TRUE"
scsi1:1.fileName = "F:\Quorum Disks\Shared.vmdk"


I'll likely do an updated post on this in a while...and will link to it once it becomes available.

Trend

Hi there,
Thank you for this very usefull post. But I still have a problem to add the second node to the cluster. I tried to select advanced then select the "Advanced (minimal) configuration, but the problem still occur.

Here are the error I see during the "Add node Wizard":
x Establishing node connection(s)
x win2003-node-2: connecting to server ...
x Error: 0x80070005: Access is denied.

In the CLCfgSrv.log, I see the following errors:

2006-08-11 10:55:20.466 [ERR ] win2003-node-1: [TaskCancelCleanup] Could not get the managed resource enumerator for node win2003-node-2.phiet-cluster.com. (hr=0x80070490, {05AA0768-5F49-49CD-AFDC-96F9D51802D4}, {00000000-0000-0000-0000-000000000000}, 1, 1, 1), (null)
2006-08-11 10:55:20.607 [ERR ] win2003-node-1: [TaskCancelCleanup] Node win2003-node-2.phiet-cluster.com cleaned up. (hr=0x80070490, {05AA0768-5F49-49CD-AFDC-96F9D51802D4}, {00000000-0000-0000-0000-000000000000}, 1, 1, 1), (null)
.
.
2006-08-11 10:58:21.279 [ERR ] win2003-node-2: (null) (hr=0x80070005, {CD36919C-9F31-46B4-A29D-AC87F4E6CC93}, {20926479-AA45-47A6-9DF7-E0144D5D03AB}, 0, 1, 1), (null)
2006-08-11 10:58:21.498 [ERR ] win2003-node-2: (null) (hr=0x80070005, {93C32F99-39CA-4D38-9D7F-2707CA0EAF46}, {E88B08DD-2460-4DF1-B5C2-28587DD164EA}, 0, 1, 1), (null)

Anyone has any idea ?
Thank you in advanced !

Trend

I was able to add the second node to the cluster now by login the domain controller and using the cluster admin to add the second node from there.

Hope it helps !
Trend

YO!

Thank allot for the tut, works perfectly. I've actually installed an exchange 2003 cluster using vmware 5.5.4 build 44386.

only thing is when adding the quorum and resource in the .vmx file, vmware automatically changes the scsi1.present = "TRUE" to FALSE!!!! and spits out about incopatibilty blah blah blah, chaged back to true and worked fine.

this might help someone who's experimenting with newer vmware.

thanks

The comments to this entry are closed.