iSCSI to HP MSA

[BobFera]

Hello,

I'm about to begin a Veeam evaluation, and I'm trying to get my would-be Veeam server configured before installing the trial software. Unfortunately I'm having a heck of a time getting Windows 2008 to see my iSCSI VMFS (HP MSA 2012i) volumes correctly. They show up in Disk Manager as "unknown/uninitialized" instead of "online/ready". I've seen these symptoms described in several posts on this forum, and I've tried to incorporate as many of the suggestions I've found as possible, but with no luck.

The volumes are presented by the MSA as "read-only", as Veeam recommends (I'd be very reluctant to do anything else in fear of Windows trampling over my VMFS volumes anyway). I've installed MPIO, and my 4 paths to 2 volumes show up as 2 drives to Windows, so that appears to be correct. I've seen references to people installing vendor-specific drivers (Dell, I believe), but there doesn't seem to be anything like that available from HP. HP support has not been helpful so far (case open for about a week).

I know this isn't a Veeam-specific question, but I was hoping that someone out there might have a hint or two that could help me get going with my trial...

Thanks very much,

Bob

[Gostev]

Hi Bob, I would recommend to try uninstalling MPIO and rebooting the server. I have seen MPIO causing issues (at least with VCB) before.

Also, to be able to start evaluation while HP is looking for solution, you could install Veeam Backup in a VM and use vStorage "Virtual Appliance" mode (be sure to configure Veeam Backup VM with 4 vCPUs).

Hope this helps!

[Michael_6835]

Hi,

I'm not sure of the specifics but I have a MSA2312i and it is whats housing my vsphere environment. I am using a windows 2008 std server as my veam backup server via the windows iscsi initiator. I have not installed any multipathing software in the vsphere environment, only using the default vmware round-robin options. (not sure it applies, but figured i would state it.) There is no software being applied on the backup server itself other than veeam install and iscsi initiator.


On the MSA side of things, I first added the windows server as a host and configured the mappings exliplicitly to be presented as read only as you have. I then used a chap secret. I am only using one spare nic in the server, so I had to make sure that the nics on the msa controllers were using the subnet that I was plugged into. (I have 2 separate iscsi subnets)

From the windows server, I went ahead and disabled the automounting option via the command line. I then configured the iscsi target for an address on the controller that I plugged the nic in. Applied the chap password and selected to have it auto start. I was able to successfully connect and view the added drives in disk manager, it should only say read only.

It took some playing around at first, but ultimately, it was something that just worked. I know i really didn't give specifics, but thats only because there wasn't many. I would double check the settings again and make sure you didn't miss something.

I can double check my config tomorrow if you need me too.
let me know, i would imagine that the msa's should not be that different in the admin page.
Mike

[Michael_6835]

i just looked at my setup.

a few things: uninstall the MPIO software.

Also in my server they do not read "online/ready" they read as a basic disk and only lists "read only"

the fact that it is showing you the it is unitialized hints to me that windows sees it as another disk that it can use. You obviously know you don't want to initialize this disk. Did you run the command prompt to disable to the automount in windows?

[BobFera]

Hi guys,

Thanks for the responses and sorry for my delayed reaction. Got dragged in a couple other directions the past few weeks...

Anyway, here's where I am.

I've uninstalled MPIO from the Windows box and cleared out all the iSCSI initiator settings -- starting fresh.

My MSA has two ports on each of two controllers (i.e a total of 4 ports/IP addresses). I have two volumes on it that are presented as Read Only to my Windows iSCSI host on all 4 ports.

At the Windows command prompt I go into Diskpart and type "automount disable". On the iSCSI initiator I add the IP address of one of the controllers' ports (i.e. only one path to the msa) on the Discovery tab. Once I do this, one entry shows up in the Targets tab. I click "log on", choose "automatically restore..." and click OK. At this point one disk shows up in Windows Disk Management with a status of Unknown/Not Initialized. And no sign of the second MSA volume (which I think I saw when I had MPIO installed and 4 "portals" defined in Discovery -- assuming of course that the two drives were really the two volumes... ).

So, I think I've now done what you guys have so graciously suggested, but I'm still not seeing what I need to see. I must have missed something... Any idea what it might be? Meanwhile I will keep staring at it...

Thanks again,

Bob

[BobFera]

One more tidbit...

I added a second "portal" to the Discovery tab, giving it the ip address of port 0 of controller B. When I did this, then a second target showed up. Logging into that gave me a second Unknkown/Not Initialized drive. It appears that one of two drives is on LUN 1 and the other is LUN 2, so I think I have a true one-to-one match between the Windows drives and the MSA volumes. So there's another mystery... Don't understand why I have to attach to both MSA controllers to see both of my volumes when the volumes are mapped to both ports on both controllers...

Finally, in Diskpart I did a "detail disk" command for each disk. Both of them say "read-only : No". But the disks are presented by the MSA as read-only...

Sigh.........

[Michael_6835]

Bob,

How did you create the Vdisks/luns?

Upon first thought from just reading your last post, it almost sounds like your vdisks have been statically assigned to a specific controller. I've read about it on some forums, it's a way to manually load balance the load. But i am no expert with sans. Have you tried contacting HP?
I would start with them your trying to get the access to windows before veam even enters the scenario.

I currently only have one vdisk, so i can not speak for certain on the manual addressing of the controller.

[BobFera]

Hi Michael,

The Vdisks are presented to both ports of both controllers. ESX has been running with them fine for months, seeing all the paths.

I've had a case open with HP for over a month. It has been 3 weeks since I've heard from them, and they continue to ignore my case updates and emails requesting status. Weird... But I will keep trying with them. I haven't even downloaded Veeam yet. As you suggested, I wanted to get the basic communication working before I started playing with the application software (and being on a trial period clock). I posted to this forum as a shot in the dark that someone else might have run into a similar situation. I must be doing something stupid...

Thanks again,

Bob

[Gostev]

Bob, can you PM me your HP support case ID number, I will talk to my HP contact and ask the status of your support case... we've been working together with HP lately on testing and certifying some HP storage solutions for Veeam.

By the way, there is no problem extending your trial period once you run out: you just talk to your Veeam sales rep and he/she will be able to get you extended trial license manually generated.

[JJones]

I don't know if it's related or not, but I had the exact same issue with my Compellent SAN. Support from Compellent and vmware and the other backup solution we were trying at the time (which can't hold water against Veeam) proved useless. What I found is that volumes formatted in VMFS version 3.33 displayed properly, but 3.31 and older would show up as unknown/uninitialzed. I took that opportunity to clean up our datastores by creating new volumes (which got formatted in 3.33) and now all is well.

I realize that there's no reason the VMFS version should make a difference, and VMWare told me the same thing, but they also didn't have any other fixes for me.

[BobFera]

Interesting. I just checked, and my VMFS volumes are 3.31. Unfortunately, if true that becomes a complete nightmare for me as all the available space on my SAN is dedicated to my existing VMFS volumes.

Thanks,

Bob

[Michael_6835]

bob,

that has to be the issue, i started my environment new with vsphere, so i wouldn't of had that issue. sorry
perhaps you can use the virtual appliance option in Veam. Maybe Gostev can highlight that option for you if it proves to be an issue.

One other option i can think of is to get some local drives and put them in the server. then migrate the machines over to them temp.

not sure if that is even an option, definitely want to consult others.

[Gostev]

Gostev did highlight the Virtual Appliance option already - in his very first response to this topic

[BobFera]

Well, not what I wanted to hear of course. Kind of surprising that an issue like this would come up, too, as this is a pretty new VMware installation (last summer) -- vSphere was bleeding edge at the time so I went with the latest available version of 3.5.

Thanks for the info. I will also take this up with VMware.

Regards,

Bob

[BobFera]

Hi,

Well after going around and around with this I finally had the brilliant idea of contacting Veeam support. Even though I hadn't even downloaded a trial copy yet I got a quick response form a support engineer who said I've been worrying about nothing all along. He said it doesn't matter how Windows displays the LUNs. As long as they are there, Veeam can read them.

This morning I downloaded a trial copy and ran a test backup, with the job set to "SAN only". Lo and behold, it worked (backup of a 16GB VM took less than 3 minutes)!

So I wanted to make one last post to this thread in hopes of preventing others from falling down the same rat hole. iSCSI SAN-only backups of ESX 3.5 (VMFS 3.31) VMs do work even though Windows Disk Management shows the LUNs as "unknown/not initialized".

Thanks again to all who offered advice. Now on to the eval...

Regards,

Bob

[ian0x0r]

I know this is an oldie, but I came across this whilst trying to fix the same problem using the opposisitions backup product....

In my case, the iSCSI initiator was set to point to the incorrect controller on the SAN, so it could see the disks, but they showed as not initialised. Changing the iSCSI initiator to point to the controller that actually owned the disks solved the issue. Disks show as online healthy and SAN backups work as expected.

[Gostev]

Thanks Ian, this is good to know.