sasar
Contributor
Contributor

Can't start my virtual machine which is part of MSCS

Jump to solution

Hello,

Recently when i shutdown one of nodes in mscs cluster. I have a problem to power it up since i have a message "Unable to access file since it is locked". I've tried to remove machine from inventory and add it again, but that didnt help. I have also tried to change ESX host , but also without success. Other machine in cluster is powered on correctly.

So anybody have an idea about this ? How to solve the problem ?

0 Kudos
1 Solution

Accepted Solutions
TheRealJason
Enthusiast
Enthusiast

I had this issue when I had DRS fully automated in my cluster. It was trying to power on the VM on the same server that the other node was on, so it was throwing that error.

I know it is not supported to run an MSCS cluster in a DRS cluster, but it does work, and it is our QA environment.

If it is fully automated, you could try changing it to partially, and when it suggests which host to power it on on, then just choose the one that does not have the other node.

View solution in original post

0 Kudos
8 Replies
jitendrakmr
Hot Shot
Hot Shot

Is it giving you details about which file is locked? Can you please attach vmware.log file of that VM?

try running "ps -aef|grep <machine_name_case_sensitive>" on that esx host and see if that vm is already running? If so then note the process id and kill it and then try restarting that VM.

Jitendra Kumar

MCSE 2003, VCP, CCNA, ITIL Foundation

VCP, MCSE 2003, MCITP Enterprise Admin, CCNA, ITIL Foundation, Netapp NS0-153 (Storage Networking) Personal Website - http://www.virtualizationtrainings.com, http://www.hillsindia.com
0 Kudos
sasar
Contributor
Contributor

No it doesnt give me any details about file. Here it is the log file of this virtual machine.

0 Kudos
Zleipner
Enthusiast
Enthusiast

Hi there,

I looked through the logfile and seems like it's having a problem opening vmdk file: "/vmfs/volumes/47a1f7f9-b7c659d2-7013-001a64353f68/test02/test02_2.vmdk"

I assume this is shared storage or Quorum disk for the cluster, if that's not the case then disregard this post. Also I assume both nodes in the cluster are virtual machines.

I would start by checking the settings for the SCSI controller for the shared storage disk and verify that it's set to SCSI Bus Sharing - Virtual and that it's an LSI Logic adapter. You cannot use the first SCSI controller as far as I know so also make sure you have a separate SCSI controller for the shared storage disks.

Let me know how this works out and we can look deeper if it doesn't help.

Cheers!

0 Kudos
sasar
Contributor
Contributor

Hello,

Yes you ar right this is shared storage and both machines are virtual in MSCS cluster. And i have separate scsi controller for shared storage disks. And it is LSI logic.

Only i use Physical compatibility since i want to run this machines on separate esx host and not on same one.

This machine worked for me till few days ago.

So what to do in this case?

0 Kudos
Zleipner
Enthusiast
Enthusiast

Hmm, I was under the impression you couldn't use vmdk files at all to cluster across boxes. Was it perhaps so that when it was working, both nodes were on the same ESX host and now they have moved to different boxes?

See table A-2 from http://www.vmware.com/pdf/vi3_301_201_mscs.pdf

Cheers!

0 Kudos
kjb007
Immortal
Immortal

Disk states RDM, so I assume you are using an RDM, and not a shared vmdk file? As stated, the shared vmdk file across servers is not supported.

-KjB

vExpert/VCP/VCAP vmwise.com / @vmwise -KjB
TheRealJason
Enthusiast
Enthusiast

I had this issue when I had DRS fully automated in my cluster. It was trying to power on the VM on the same server that the other node was on, so it was throwing that error.

I know it is not supported to run an MSCS cluster in a DRS cluster, but it does work, and it is our QA environment.

If it is fully automated, you could try changing it to partially, and when it suggests which host to power it on on, then just choose the one that does not have the other node.

View solution in original post

0 Kudos
sasar
Contributor
Contributor

Thanks mate,

You were right it was DRS cluster fully automated. When i changed it to manual mode , i've got option to choose on which host should be machine powered on. So that was the reason for this mess.

Thanks again for your helpfull answer.

Cheers

0 Kudos