Sharing RDM Disks for MSCS

[Heartstealer]

Hi Guys,

I guess that I will need to share RDMs between VMs if I need them to be a part of MSCS Cluster. How would I do that ? Does anybody have any good documentation for doing a MSCS on 2 VMs on same ESX and on different ESX? Also what are the best practices to be followed for MSCS on ESX?

Thanks

Raul

[whynotq]

Hi Raul, well your timing couldn't be better. just going through this scenario and all the pain it has caused me due to my own lack of knowledge on the matter. But hey, this is how we learn so,

MSCS with RDM. Import the RDM to a seperate folder to the nodes. the SCSI controllers must be shared. The Disks should be in physical mode. you cannot use VMotion. You cannot have the MSCS nodes on ESX hosts in a cluster with HA or DRS enabled. then there are a few advanced settings under the disk options. i'll attach the pdf docs i have found on the topic.

good luck

Paul

[Lightbulb]

Try the following doc for cluster in a box http://exchangeexchange.com/blogs/bkeane/archive/2007/07/30/mscs-clustering-in-vmware.aspx

Same procedure works across hosts you just need to use RDM presented to the VMs "physical compatibility mode"

[paulmack]

I'm running a 2 node SQL 2005 cluster on hosts that are in a DRS/HA cluster. To get round the Vmotion issue I've disabled DRS/HA for the clustered SQL VM's. Not had any issues with this setup and we've been running it like this for about 4 months.

Cheers

Paul

[Texiwill]

Hello,

Moved to VIrtual Machine and Guest OS forum...

Check out http://www.vmware.com/pdf/vi3_35/esx_3/vi3_35_25_u1_mscs.pdf from VMware on how to do this.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

Blue Gears and SearchVMware Pro Blogs: http://www.astroarch.com/wiki/index.php/Blog_Roll

Top Virtualization Security Links: http://www.astroarch.com/wiki/index.php/Top_Virtualization_Security_Links

[atbnet]

Use the VMware docs for a detailed view and once you know whats what this is how to add and RDM, simple process:

http://www.vmadmin.co.uk/index.php/resources/35-esxserver/58-rdmvm

Andy, VMware Certified Professional (VCP),

If you found this information useful please award points using the buttons at the top of the page accordingly.

[Heartstealer]

Paul,

Thanks for your time and effort.

However, my client is looking at MSCS on a HA/DRS Environment. RDM mapping can work in two methos right Physical and Virtual so if I map it by virtual method I will be able to make use of DRS and HA am I heading in the right direction guys?

I want an MSCS cluster running in a HA/DRS environment

Thx a ton guys

Raul

[kjb007]

Virtual mode bus sharing can not be done with RDM's when your vm's are on two different hosts. Virtual mode is for sharing devices on the same ESX host. Physical mode is for sharing across hosts. Bus sharing itself precludes you from vmotion, so while your vm's can be in a HA/DRS environment, the vm's will not vmotion. HA will work just the same, only vmotion will cease to be an option.

-KjB.

[whynotq]

The documentation says it is not supported in HA/DRS, it took 5 weeks with an open Sev 2 case before i had some one tell me this, i have turned off HA/DRS in the cluster but we are working around as HA is a requirement for the project. I have found that the VMs run but it is from the ESX level that we see issues:

1. massive amount of SCSI reservations

2. hosts take 10 - 15 minutes for vmhba rescan

3. datastore browser slow to respond

4. Host can take 10 minutes to shutdown and then 30 minutes to reboot

if we take the MSCS shared away from a host the host responds as it should. Basically, in a small setup i.e. couple of hosts and few RDMs you'll probably see no issues but in a larger environment, we have 4 hosts dual attached and 22 volumes including 8 RDMs, you may see similar issues as well as VMware support will say it's an unsupported configuration.

[CLMILLER]

I have had similar issues in my environment. Not only is it unresponsive, I have noticed some wierd behavior when adding new luns into the environment.

1. Cannot add new luns into the environment unless the rescan takes place on a host that contains one of the mscs servers.

2. I have noticed that luns that are removed from the environment will remain until it is scanned from a host that has one of the mscs servers

3. Every vmkernel log is bombarded with SCSI errors that seem irrelevant (obviously the aren't)

4. Overall guest performance has significantly been affected since MSCS implementation (Monitoring 20 host in the environment with opsmgr 2007, performance metrics took a significant hit starting the day of implementation, most noticeable is CPU)

When I removed the MSCS from the HA/DRS cluster all problems stopped and performance issues ended.

Environment notes: ESX 3.5 U3, HA/DRS, 2 MSCS SQL 2005 Servers w/ 4 shared physical RDM's, FCSAN (EMC cx700), MRU pathing, 16 host and 96 guest, BL480c's....

[Heartstealer]

As you guys said ... I am also experincing problems ... I havent installed an MSCS service ... i just have a file server with 16 LUNs of 128 making of 2 TB and we are analyzing the performance and then thinking about clustering the same if required... I have the problems of SCSI messages on my console of the server ... obviously it will be thr in the logs ... my HBA scan takes ages ... Is VMDK better than RDM Drives from experince ... in most of the documentation it says performance is not at all an issue.. I just wanna understand ... We have a stage 2 of the migration when we will convert the current connected RDM Drives to VMDKs usinng a VMWare converter ... I hope that shouldnt be a problem.