Curses!

OpenFiler was originally introduced to me by Greg Shields (ps, that article is a good read) as part of his training series on VMWare ESX Server, where he demonstrates the VMotion Technology using an OpenFiler IP SAN.  This is what originally gravitated me towards using OpenFiler in my own lab.

Well, I got all of my resources online and ready to go.  I added the Failover feature to each node of the cluster.  I started by creating a new cluster using just a single node.   And then when I tried to add the 2nd node, the cluster validation wizard was showing failed checks related to the shared storage.  I should have taken a screen shot of the exact text.  But the point is, while searching for what might be the cause of the problem, where should I find the answer?  Why Greg Shields of course.  This time in the form of an entry on his Blog entitled “No, OpenFiler Does Not Work with Windows Server 2008 Clustering“.

Apparently there are a few solutions out there that will work, however they are either not free, or are only provided by Microsoft OEM providers in the form of a specialized version of Windows Server called, “Windows Unified Data Storage Server 2003″.

The cause of the problem is that Windows Server 2008 clustering requires support of persistent reservations.  I’d have to say that I’m a fan of persistent reservations, they can prevent some nasty data corruptions that can occur in a shared storage environment, but dang…  I wish I could disable the requirement for my testing.  :(

blog, Cluster, iSCSI, Microsoft, Openfiler, SAN, training, vmware

First, if you haven’t already read completely through my last post, the rest of this post won’t make a whole lot of sense.  In summary, I had complained about the performance I was getting using iSCSI on my network.

I noticed this on the Openfiler administration page today.

Notice the highlighted drop packets on the eth1.

I bonded the two NICs together, and configured them for load balancing, but apparently eth1 is having some problems.  This is probably due to the NICs not being identical.  I’ve gone ahead and removed eth1 from the bond.  Hopefully when I configure the SQL Cluster today, I’ll have better things to say about the iSCSI through-put.

Cluster, iSCSI, Openfiler

I wasn’t very happy with the quality of my last video. I had fun making it, just didn’t like the way Youtube compressed it.
I’m experimenting with trying to host my own videos locally and we’ll see how this turns out. Hopefully the screen text will be much sharper.

The big picture is that I’m wanting to play with the failover and nlb capabilities of Windows Server 2008. I currently have my Hyper-V box online, as well as a domain controller for the march.local domain. I also have a 2nd physical box that I dumped every spare IDE drive I could find lying around into, and installed an opersoure iSCSI Target called OpenFiler. This is really my first time playing with iSCSI or a linux based SMB file server, so it’s been a unique learning experience.

I had already installed Openfiler, and after a little tinkering, I got kerberos authentication working and joined the box to the domain. At that point I brought up my 3rd Windows 2008 server and got an itch to play with Windows Deployment Services. It was the one thing I didn’t experiment with live while studying for the Windows 2008 Server exams. As I was walking through the wizard it prompted me for a place to store all of my images, and rather than creating another .VHD file to store them on a 2nd drive locally, I opted to go for the iSCSI option.

So here is a quick video on setting up iSCSI.

So since setting up iSCSI, I’ve got WDS working. I was able to use a boot.wim to PXE boot a Virtual machine with an attached Legacy adapter and install Windows 2008. After I got the new OS up and running, configured and patched. I was able to then create a capture image on the WDS server, run SYSPREP on the base OS, and PXE boot the capture image to create a new Install image to be used for future deployments. With that said, I will not be deploying using the image I just created.

It was a painfully slow process pulling the image from the iSCSI storage. No doubt because I have 3 Virtual machines all attached to the same physical network adapter, which is also being used to direct the iSCSI traffic. *UPDATE HERE* I may try putting Netmon on the Parent OS and see if I can monitor the network traffic and see exactly how bad the bottle neck is. It also may have something to do with the legacy network adapter booting at fast ethernet speeds, and not properly negotiating it’s duplex settings, since the physical network is all gigabit. I’m not completely clear on how a Virtual Switch would handle that. Getting the network properly segmented is definitely on my to-do list. I have 4 physical adapters in the box, so if I had another switch, or even a layer 2 switch, I would be able to properly subnet my network out.

So, for the future builds, it’s much easier to just run SYSPREP with the shutdown command, and then copy and rename the .VHD file 5 more times.

Oh, and one last note on Windows Deployment Services. I had my server set to only accept known clients, which means you have to pre-stage the machines in Active Directory. The directions to do this are:

To prestage client computers

Yeah, ummmmm, no next button here......

Hyper-V, iSCSI, NLB, Openfiler, Video, wds, Windows Deployment Services