iSCSI repository setup issue

[barchas]

Hi Guys.

I have searched hi and low for a solution and I must be missing a step that no one ever mentions, or I am being dumb.

My environment:
VBR 6.5 on virtual Win2008r2
Vmware 5.0u1
repository is freenas9, presenting an iSCSI target with no restrictions or permissions.

I am able to mount the iSCSI target on all ESXi hosts, and did performance testing against it with my XP VM that has IOmeter on it.
No issues there.

I am able to attach it to windows via windows iSCSI initiator. It sees the volume, and it does show up in Windows Disk Manager.

However, it is never actually available as a target in Veeam.
In Veeam I go to the Backup Infrastructure tab.
Then Backup Repositories.
Then click Add Repository.
give it a name, next.
Select Windows Server, next.
then on the Repository Server page, "this server" is selected, and I click populate. All I get back is the C drive. Because obviously there is no Drive letter assigned to the iSCSI drive.

I have tried bringing the disk online/offline, and no joy.
Cant "browse" for it on the next page either.

All the instruction I find just kind of gloss over this part, focusing more on the actually windows mounting of the iSCSI target, which of unfortunately is the one part I didn't actually need help with... It sounds to me like everyone says "it should just magically know to use the iscsi target and work", but obviously that is not the case, nor is that what I actually think anyone means. But the lack of info on this step is a bit baffling. Unless of course, I have some magic situation that is screwing me up somehow?

Any help? Saturday night beer has not enlightened me either...

Thanks!
Dave

[kmount]

Hi Dave,

If you format the drive from your Veeam server to NTFS it'll get a drive letter and show up in the list.

Veeam (or anything else) can't write straight to iSCSI, it needs to be formatted with a filesystem first, be that VMFS for VMware, or NTFS for Windows.

So hopefully, if you can 'see' it in disk manager which it sounds like you can if you can online/offline it, then format it and it should show up and let you carry on.

Hope this helps.

Kim

[barchas]

Thanks for the info Kim.
So, in all the tutorials and how-tos and Veeam docs, this is NEVER mentioned... In fact they all hint at that you should never even initialize the disk because it will kill any existing VMs on it, which of course makes sense. In my case, this is a new array so no risk there and don't even care if the whole thing gets nuked and I need to rebuild. But I am still stuck on the fact that everything I read says not to do this. lol.

So if you don't mind a followup then.
Say i have a big giant single LUN. 30t in this case. Aside from the arguments about LUN sizes so someone doesn't come in and derail with some pointless one sided lecture about it... If I have this big LUN, but want to have both VMs running from it AND backups going to it, this now sounds impossible. Mind you I can obviously carve up the LUNs to have a portion for VMs and a portion for Veeam repo, but now I am into the realm of trying to understand the utter lack of mention of this in all the stuff I read. Because this now makes me wonder how people are backing up from iSCSI targets? Because all the mentions of not initializing I gather have to do with people backing up existing VMs from iSCSI. Which now means, do I need to have both the source array and target array visible in Veeam? I didn't think I did because I use virtual veeam proxies. Im just now a bit more confused i think.

And again Kim, thanks for the initial info. I most likely am going to do exactly that, format NTFS and be done with it. Ive wasted more than enough time on this already and just need to get moving forward.

Dave

[barchas]

Update. Now that I have it setup to be NTFS formatted, I can select it and it works technically.
However, it is not doing Direct SAN it seems. And that would make sense since I think veeam thinks it is a local disk and has no idea it is iSCSI target. SO, I don't think formatting as ntfs and giving it a drive letter actually got me any closer to Direct SAN which in the end is the real goal and why I was trying to get Veeam to see the iSCSI disk.

Any other thoughts?

[yizhar]

Hi.

I think that best course for you would be to contact support, so they can guide you to solve basic issues.

In general I would like to emphasize some conceptual topics:

* You should not mix source datastore with target repositories.

The first - source datastore is formatted with VMFS and is where your VMs reside.
Veeam needs to read from it, either with direct SAN access, or via the esxi host (proxy mode or NBD).

The second - target repository, is where backup VBK files are stored.
It is formatted with NTFS, and Veeam needs read+write access to it.

Source and Target are different things, and they should never be the same ISCSI volume.
Also - you should never store your backups on the same device (wheter NAS or SAN) as the source. Think what can happen if the device fails, either by hardware issue, human error, theft, you name it...

So - for direct SAN access, the basic design is as follows:

Source datastore - needs to be accessible (read only) to Veeam server. In disk management should be in "offline" status.

Target repository -
Needs to be on separate device. If you planned to store both VMs and their backup on the same FreeNAS device - you need to replan. For example you can use local SATA disks on the backup server.
Then make sure that Veeam server has drive letter + read/write access to the repository.

Again - I suggest that you also contact support, so they can help you by phone, answer your other questions and guide you further.

Good Luck!
Yizhar

[Vitaliy S.]

Update. Now that I have it setup to be NTFS formatted, I can select it and it works technically.
However, it is not doing Direct SAN it seems. And that would make sense since I think veeam thinks it is a local disk and has no idea it is iSCSI target. SO, I don't think formatting as ntfs and giving it a drive letter actually got me any closer to Direct SAN which in the end is the real goal and why I was trying to get Veeam to see the iSCSI disk.

Can you please elaborate on this - " it is not doing Direct SAN it seems"? If you format your SAN volume with NTFS and then present it to the backup server, all VM backup traffic should go via iSCSI network to the target LUN.

[barchas]

I think I may be over interpreting the word "direct".

I realized that of course traffic still has to flow through the Veeam server/proxy. since its doing dedupe and compression.
So "direct san" just means that veeam speaks directly to the san, not, it facilitates San to San.
Source San -> veeam server -> Target San.
Traffic is going over 10g storage net, I just for some reason firmly believed that it would skip the Veeam server. Like agent to agent. But apparently when you stare at the trees for too long you can't see the forest.

I got it all working now. Here is my short description of steps, just worded differently in case some other "over thinker" shows up...
Target SAN is connected via Windows iSCSI, initialized and then formatted to NTFS, then assigned a drive letter in windows drive manager. Now you can see it as a drive in veeam (and windows for that matter).
Source SAN is just connected to via Windows iSCSI, it is NOT initialized or brought online in windows drive manager, and nothing is done in veeam to see this. It just automagically works.
Confirm by seeing traffic traveling over storagenet during a backup.
Confirm in veeam during backup by seeing how it mounted the VM disk (hot add, LAN, or SAN) of the VM currently being backed up with SAN being a win obviously.

Thanks guys, this was a case of having my head firmly up my A$$. Hard to see in that case...

[Vitaliy S.]

Thanks for clarifications and sharing your setup config with the community members.