iSCSI targets in guest vm. How do I get them backed up?

[mascaroit]

I am evaluating Veeam B&R v6. I have a few virtual servers that have iSCSI targets for data. I have read several mentions about mounting them as vRDM's and then Cold vStorage Motioning them to another datastore as that will convert them to VMDK's and then they will be visible to Veeam.

Does anyone know of a rock-solid walkthrough to accomplish this?

Thanks,

Tony

[foggy]

Tony, you can simply map the LUN to guest using virtual RDM mode, no need to convert it to VMDK actually as Veeam B&R can correctly process vRDM disks (moreover, it will convert vRDMs to VMDK itself, during backup).

[mascaroit]

Thanks for your quick reply.

I must be missing a step. Do you have any refernce material that can step me through the process? This is production data and I cannot afford to goof up.

[Gostev]

If you are talking about setting up vRDM disks, you should refer to VMware documentation (admin guide).

[mascaroit]

OK, here is what I did:
Disconnected the volume from the iSCSI initiator within the guest OS. Set the iSCSI initator service to manual.
Shut the virtual machine down.
Edited the virtual machine's settings to add a vRDM mapping to the data volume on my SAN.
Booted the virtual machine. The drive and data appeared as expected and not through the iSCSI initiator. WIN!

Then, I opened Veeam B&R v6 and added that virtual machine to a backup job. The vmdk for the guest OS was discovered, but not the vRDM.
Going by what Foggy posted earlier, if I run a backup of the virtual machine, Veeam would somehow convert this vRDM to a vmdk even though it does not appear in the backup selection, is this correct?

[Gostev]

Not sure what you mean by "backup selection", as there is no place in backup job wizard where we show individual disks for the VM.
Yes, during the backup your vRDM disk should get written into the backup file as a VMDK file.

[mascaroit]

OK. I ran a backup and the data was written to the backup. However, in order to get the job to process correctly, I had to create a volume on my Equalogic SAN slightly bigger than the one I was backing up and have the vRDM mapping pointing to this empty volume. I do not get it.
Is Veeam storing the snapshot in the empty volume?
Now I am occupying over a Terabyte of my SAN just to back up a 510GB volume.
I guess I am totally lost now.

[Gostev]

You really want to check out VMware admin guide, or you will remain lost forever

VMware snapshots do require a place to store snapshot file, this is where any writes are redirected while the snapshot exist. Since the snapshot is only present for the duration of backup (minutes/hours), the disk space required is usually just a few GB (just enough to host all blocks changed in the virtual disk for the duration of backup).

[mascaroit]

So instead of having a greater than or equal to volume, I should just create a small 4 or 5 GB volume to store the snapshot data during backup?
I have read the storage section of the admin guide and it is very vague about how to accomplish what I am trying to do.
I cannot be the only guy here that is in the same boat.

[Gostev]

You can just use the same datastore where VMX for this VM resides, sure it has a few GB of free disk space?

[mascaroit]

The volume where the vmx for the vm only has 53GB of free space.
Initially I added the vRDM and told it to "Store with the virtual machine" Then ran the backup. It failed. The following is the entry in Enterprise Manager: 2012-04-03 08:47:12Error: File is larger than the maximum size supported by datastore So that tells me I need more than 53GB to store the snapshot blocks. This is the first time I have backed up this vm and data in Veeam, so I know that it is a full backup.
Now, if I run another backup and leave that 520GB empty volume present to process the snapshot, subsequent backups are just changed blocks which should not be much. That leads me to believe that I would be able to discard the 520GB volume after the first full backup has run. Then I could re-map the vRDM to the datastore that the vmx resides in. Am I on the right track here?

Thanks so much for all your help Gostev

[Gostev]

This error has nothing to deal with the free space, but with the block size the VMX datastore is formatted with. Just search this forum for the error you are getting, there are plenty (hundreds?) of posts regarding it. Or better yet, upgrade to ESXi 5 and VMFS5, which does not have those stupid variable block sizes at all.

[Gostev]

And even better, get rid of your vRDM disk altogether, and use VMDK - and you will thank me later. There are literally no benefits from using vRDM these days, however there is a large amount of drawbacks.

[mascaroit]

I am already on ESXi 5 but we did not upgrade the datastores to VMFS 5 because it seemed very risky. It seems that upgraded VMFS 5 volumes retain the block sizes (1MB) that are currently on the volumes where our VMs are stored.
So in reality, should I create new volumes with VMFS 5 and just vStorage Motion the VMs? Seems there is less risk in going that route.

I cannot say how much I appreciate your help. I have got to be trying your patience. The reason I wanted to use vRDM was there is no real way to get the data that is being stored via iSCSI initiator in the vm into a vmdk other than robocopy as far as i know.

[Gostev]

You are right, just upgrading your existing datastores to VMFS5 will not cut it, creating new datastores from scratch is what you want to do.

So in reality, should I create new volumes with VMFS 5 and just vStorage Motion the VMs? Seems there is less risk in going that route.

Yes, as far as I heard that's how everyone else are doing this. Well - those who have Storage VMotion licensed.

[largechicken]

Hello,

I have a few Windows cluster servers for redundancy in my multi site VMware setup. These have separate iSCSI targets to separate datastores for SQL clustering. Whats the best way to incorporate the full VM with iSCSI target data into a Veeam solution?

Thanks

[Vitaliy S.]

Hi Shane,

Disks connected via in-guest iSCSI initiator cannot be backed up since these disks are not visible in VM configuration file, so you should convert these disks into vRDM or regular VMDKs to back them up.

Thanks!

[largechicken]

Thanks for the reply.
Do you know if converting to a vRDM would effect performance on a SQL server cluster? Naturally i would have to take down a production cluster to make the changes so want to make sure that it's the correct path to follow. Ideally with a cluster setup 2 or more servers must have access to the same shared disks.
Thanks

[dellock6]

Hi Shane,
there has always been an endless debate about RDM/VMDK performance comparisons, but since vSphere 5.0 there is really no performance difference between them. The choice for RDM only relates to design consideration and specific needs like Microsoft Clustering, otherwise there is really no more excuse to not use VMDK.

Also, you can convert pRDM to vRDM with only a quick shutdown of the VM, and then using storage vmotion convert vRDM to VMDK, so the effort is low...

Luca.

[robbyde]

Hi All,

I've got a file server that has a a drive directly mapped to it via the isci initiator that I'm keen to backup. How do I go about this as current Veeam ignores the drive complete.

Would I have to add the drive to the Veeam server?

Thanks

[foggy]

Rob, as it is stated in Veeam B&R system requirements, disks connected via in-guest iSCSI initiator are not supported and are automatically skipped from processing. Thanks.

[dbarroco]

Hello,

I couldn't quite predict what would be the behaviour on this scenario. This is held up for a file server migration.

Original server: phisical machine with mounted iSCSI device

new VM server: 2 vmdisks - one for system (50g) and one for file shares (2T) (sda and sdb)

for migration purposes I attached a vmnic to the original iSCSI storage so that the copy would be lan free and phased, that is presented as sdc to the system.

I did include this machine on the backup job until the migration in concluded and the iSCSI device is unmounted. I could not see how a backup would behave beginning with the snapshot. since this is a production scenario could not even test. However is was curious on what would happen. Can anyone break it down to me?

Thank you
David

[foggy]

David, the VM will be backed up without the iSCSI disk, as it will be skipped during processing.

[MJeB]

Hello,

I'am evaluating Veeam B&R 7 as a backup and DR environment for our datacenters and stumbled across a question: If I backup a VM that has a LUN attached that is used inside the Machine via iSCSI connector, is there a way that the mapped LUN will be backuped along with the machine or do we have to backup the LUN through our storage appliances?

[dspjones]

I understand that iSCSI presented disks are skipped during backup but it seems to me that since MANY IT shops use such a configuration that at some point either VMWare or Veeam should figure out a way to allow this data to be visible to the backup. So the question is - will there ever be a way to natively backup iscsi targets with veeam? I'd assume the answer is no as long as VMWare can't "See" the target in the VMX but are you aware of any work on their part to make that happen? Right now, we just don't worry about the iscsi targets for backup through veeam. Since that data is on the SAN, we do replication via the SAN to our DR site. There is also a couple of iscsi targets on our production NAS but the data is "losable" meaning if the building explodes tomorrow, we could get the business back up without this data so we just ignore it for backup. But it sure would be nice to be able to back up iscsi targets natively with VMWare/Veeam.

[Gostev]

What is the point of using iSCSI presented disks in VMs in 2014?

[vmNik]

Hello all.

From what I know and seen in the past, Veeam was unable to backup a VM that contains internal MS iSCSI initiator LUN's. Is this still the case for Veeam 8 and all of the updates to-date? Is there a way to map an inside iSCSI connection to the Veeam job to perform such a task? I'm looking at a few Exchange boxes from back-in-the-day, and they were all set up this way, so I'm trying to capture these VM's with Veeam before detaching the internal MS iSCSI LUN's and going by-way of attaching them to the VMware hosts for iSCSI presentation. Doing some cleanup.

Any thoughts, please let me know. Thanks.

[foggy]

Yes, this is still the case for v8 with the latest updates. Please see above for some hints.