Recovery of existing VM storages

[rali]

Hi,

I have a system consisting of ESX-3.5-servers connected to SANs using iSCSI. All VMs and also regular backups are located in this storage.

As long as the system works fine, it's no problem to add a new server to the cluster and use the existing LUNs.

Now I'm thinking about the catastrophe case, in which all ESX-servers crash more or less at the same time for some reason (I've seen such a case before...). All the VMs and all the backups are still on the LUNs. But it doesn't seem to be possible to reconnect a new server to the existing vmfs-storages. All I can find in the documentation is the standard way to add new disk/LUNs. But this means a format of the device and thus the loss of all data.

Does anybody know a way of recovering the VMs that are located on an iSCSI-SAN? This should absolutely be possible. Otherwise, I would have to rethink the whole backup and restore scenario.

Any help would be greatly appreciated.

Ralf

[vlchild]

You would simply reattach an eSX 3.5 host to the storage once rebuilt, or whatever, then browse the datastore (by right clicking on the data store name). ESX3 keeps all respective files in a directory named after each machine. Browse to the .vmx file, right click, and say 'Add to Inventory'.

This is all I had to do when I rebuilt the ESX host.

Lee

[thickclouds]

I agree with vlchild.

[rali]

I'm afraid, my question was a bit unclear, because I meant

something different.

I will try to clarify my problem with a simple example: Let's assume, I have just one ESX-server connected to one

SAN via iSCSI. This server crashes completely and is irrecoverable. The SAN and

all the data files (i.e. my VMs) are still OK. Now, I have to buy a new server

and install VMware from scratch.

My question is simply: How can I use the existing LUNs

with my new server without having to destroy the data?

In the VMware documentation there is only described

how to add a new LUN to the system in a way that the LUN is formatted with

vmfs. But I don't want to format the LUN. I want to recover all my VMs...

I simulated the catastrophe case by just removing a LUN with unimportant data from the collection of storages. Now I want to reconnect it. In order to browse the VM-files I have to add the LUN. But this means formatting it!

I cannot "browse the datastore (by right clicking on the data store name)" as vlchild proposed because at this point there is no datastore name.

There should be a way of adding an existing LUN to the system without destroying it. At least it should be possible to recover the data by using a mount tool or something similar.

Thanks for your answers!

Ralf

[James-D]

You will need to build your new ESX server install the licences. Then go to stoirage and click on the iSCSI adapter enable it. Add the address of the SAN. Copy the iSCSI initiator and add it to the LUN target on the SAN. Click Rescan on the ESX server and the LUNS should be shown. The datastore will then be shown and you can add the VMs to the inventory as said above.

James

[rali]

In my test case I removed a LUN from the Storage. This is my simulation for a total loss of the ESX-server(s).

Then I go to "Storage Adapters" and click rescan as you proposed. Of course I can see the LUN under SCSI targets. But to really use it again I have to go to "Storage" and click "Add storage", which will format the LUN.

Is there something I completely misunderstand? Does the system behave differently after a fresh installation?

I know how to add VMs to the inventory. But to do this there has to be a storage defined.

Ralf

[kesparlat]

Install a new server, apply patches, etc... And simply rescan the HBA, you will see again your datastore.

Maybe if you're worried about that you should use VCB, and take your VM backups to a tape.

[bfent]

Once you've done as James' suggested, you would right-click on the datastore and select 'Browse Datastore'. You would then navigate to the VM's folder and right-click on the .vmx file and select 'Add to Inventory'.

[rapier]

I understand the issue rali is facing. I just came from our test lab working on DR strategies. I think you guys are missing a step. Once you present the LUNs to the new server the first step is to rescan the HBA's. At this point they show up under the "Storage Adapters" section. However, the LUN is not available to browse nor visible under the "Storage" section until you finish the "Add Storage" step. This is where the problem occurs that rali is referring and my problem as well.

When you go to "Add Storage", you are presented with any LUN's not previously registered. If you select the LUN with contains the data that you want to recover intact, as far as I can see it leaves you with only one option; reformat.

Does anyone know how to "add storage" without invoking a reformat?

[mcowger · Accepted solution]

Yes - rescan the storage is the step that is needed - no need to 'add storage'. If you dont see it, investigate the AllowSnapshotLuns option, then rescan again. With any FC?iSCSI lun you do NOT need the add new storage option.

--Matt

[rali]

rapier is right. That's exactly my problem. I asked VM support and they told me to follow the procedure described in http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=648264....

That's probably what mcowger meant?

Ralf

[Paul_FJS]

I am facing exactly the same problem with my iSCSI storage. I know the LUNs are fully functional and tha data is all in one piece but I cannot reconnect to the LUNs without clicking on 'Add Storage'. But I know doing this WILL destroy all my data. My LUNs are visible and are being discovered by the ESX servers but they will not attached to my servers simply by clicking 'Rescan'.

I followed rali's link to the KB article that explains setting LVM.DisallowSnapshotLun to 0, but this didn't help at all and I still cannot access my VMs on my iSCSI LUN.

I would appreciate it if someone could help out here please as this is going to sway my opinion of using VMware in a large scale enterprise environment.

P.S.: I've tried this KB method mentioned by rali from ESX 3.0.1 and 3.5 servers.

Many thanks in advance.

[mpopoola]

Did you find a satisfactory solution to this problem?