HA storage with L2 cache and Invalid partner info error

[SynthosKralupy]

Hello,

I have troubles to create HA storage on 2-node Starwind cluster. I read technical paper at http://www.starwindsoftware.com/styles- ... Sphere.pdf.

I followed doc exactly step-by-step, but I never succeeded at last step: creating replica on second node. I always got error "Invalid partner info".

sw-create-replica2.PNG (171.72 KiB) Viewed 20738 times

Only way how to create HA storage is from cluster wizard. But it never create L2 cache.

Do you have someone idea what's wrong? My primary goal is to have HA with L2 cache.

My configuration: 2 indentical HW boxes with VMware ESXi hypervisors, Starwind version 8.0 build 7145 running in Windows server 2012 R2 virtual machine.



Thank you.

[thefinkster]

Here was the procedure we were given to create the HA w/ L2 cache for version 8 build 6884 or build 7185.

Create the storage on Node 1 (we use Add Device (Advanced)). (Do not use the Clustering side of StarWind). Choose your L2 cache there as normal (even LSFS options on thin provisioning if you want).

At this point, you'll want to create the folder structure on Node 2 identical to Node 1 if you're not using the default configuration. Then create the device/target on Node 2 as you did on Node 1, but when you get to the step where you Create the new Target, choose "Unassigned" from the drop-down box.

You'll then go to the Image file under your target on Node 1, and go into the replication to add a new synchronous replication partner. You'll enter (one of) your second node IP and select a current image file instead of creating a new device. Finish that out and you should see the image file you created to select. Choose it and finish out the configuration. The synchronization should happen once it finishes attaching the partner (a few seconds).

Hope this helps and I didn't get any of the steps wrong.

[SynthosKralupy]

Hello thefinkster,

thanks for guide. I tried it but finshed on selection existing partner header file. It said (really non-sense from my point of view) message that file already exists. See picture.

sw-create-replica-use-existing.PNG (40.73 KiB) Viewed 20731 times

BTW, are you really have build 7185? Or typo error?


Regards.

[thefinkster]

Crap Typo! Build 7145!

When you're going to create your second node device/target, do not actually Create the Device.

Choose Unassigned from the dropdown.

StarWind-HA-L2.png (17.57 KiB) Viewed 20729 times

Then create the Replication from the Replication Manager. When you add your other Node, you'll choose Existing Target instead of Create new Target.

StarWind-HA-L2-2.png (12.83 KiB) Viewed 20729 times

At which point on the next screen, you'll see a list of unassigned targets to pick from (I only do this one at a time, since the generic names they use suck, and a lot of times I'm creating identically sized storage for OS drives and Log drives, etc).

Hope this helps.

[thefinkster]

Oh yes. Just learned something too in the last few days. When you create the HA device, there will be a second SWDSK configuration file created. This file is created in the default folder that's configured in StarWind, and NOT into the configured folder you choose for the location of the IMG file.

So when you create an Image, it creates both an Image SWDSK configuration file, and an .IMG file in the folder you specify when creating the Device/Target(image).
When you create the HA replication (Replication Manager), it creates what's called an "HA SWDSK file" that partners with the Image SWDSK file. This file is located in the default storage path.

You need BOTH files...

Also, a small bug if you don't add all your replication interfaces on the initial HA creation. Adding new interfaces for replication (to go from 2 x 10Gb to 4 x 10Gb for example) will NOT save in the HA SWDSK files, so upon a restart of that particular node, the replication interfaces added after the fact will disappear. Manually editing the SWDSK file is the only fix for right now (or re-establishing the partner interfaces after every reboot, per image!)

Just a heads up if you start setting up this for production but your network configuration is not yet completely setup/purchased. Don't think it's easy to add (well it's easy to add) but it won't keep.

[SynthosKralupy]

Thanks for tips, but unfortunatelly I still cannot complete Add replica wizard. Still same problem, I can't pass Partner device setup step at all

It's paradox that wizard says "Specified file already exists" although I choosed Select existing device in previous step. For this configuration I first removed everything: cluster, devices and targets from both servers and deleted all image files. I also prepared images as you wrote: configured and connected device with L2 cache on first node and Unassigned on second node.

After that I tested different combination, but without luck again.

So I have no clue what to do now except full reinstallation Windows servers and start again.


Thanks for your help again.

[SynthosKralupy]

Wow! I found way how to create HA
I had to type any (non-existing) file name in Partner device setup dialog. Then I was able to proceed to next step, choose prepared unassigned device and succesfully complete wizard. Ufff.

I suppose it is bug, because it doesn't make sense... What do you think, Starwind?

[thefinkster]

Glad you figured it out. I was naming my image files specifically (at least until the replication creation at which point it defaulted to an HAImageX format), so I never ran into the file-already-exists-in-the-file-system-but-not-in-the-configuration issue (which is what I'm assuming).

[Anatoly (staff)]

Well, initially that was a minor bug of the earlier versions of StarWind and we fixed it in the build 7145.

[SynthosKralupy]

Anatoly, I'm not sure about this, I'm using build 7145... On both nodes.

[Anatoly (staff)]

Hmm... weird... I`ll pass this to R&D.

[jariangibson]

Seeing same issue on 7145

[Anatoly (staff)]

Yeah, we got this. As I mentioned above - passed to R&D.

[epalombizio]

Hi,
Was this ever resolved? I just setup a test setup of v8 on two physical machines and when I attempt to setup HA, I receive: Device Initialization for second partner node failed Failed: invalid Partner Info.

I'm NOT using L2 cache. I followed the quick start guide exactly.. Can't seem to get around it.

Seems like a very obvious problem that should have been caught with basic testing.

I'd like to get moved up to v8 from v6, but problems performing basic operations like this concern me.

Thanks

[Anatoly (staff)]

Hi guys,

Just a quick check - can you confirm that you are using the latest StarWind build? Current (V8) build is 7354 (5 November 2014)