Here at GCA we run classes for multiple vendors. I was tasked with creating the class build for the Symantec Netbackup 6.5 for Windows class. The hardware we are using, however, is all virtual. We are using a Netapp VTL and VMWare ESX Server in order to create the student systems. The hardware we are virtualizing on lives at our Tampa office while the class is going to be run from our Seattle office.
The build starts out pretty simple. I spun up some instances of Windows server 2003, patched them up, and made them ready to be cloned for student systems. On the netapp VTL, I created a bunch of Robots and Tape devices, which are emulating a StorageTek T9840C. I added 16 tapes, 8 on one serial format and 8 on another (as requested by Symantec). I then dropped a couple HBA's into the ESX Server and connected them to the VTL (we had plenty of HBA's laying around so the Multipathing and redundancy is nice to have seeing as how we are going to share the HBA's for all of the systems).
There were a few things that threw me. The first one was how to match up the Robot and Tape drives on each machine. When I connected the HBA's and added a new SCSI device to the VM, the pulldown list just had them by name, so I saw 8 Robots and 16 drives with no information regarding which was which. They were also not in any specific order.
To figure this out, I SSH'd over to the service console of the ESX Server and did an ls -al on the devices directory that contained all of the targets. This matched a sym link that had the LUN target directly to a ridculously long unique number identifier. This long number also shows after you add the SCSI device to vmware if you hover over it in the edit settings window.
The next thing I tripped over a bit was making it to where two virtual machines had access to the same SCSI device. The way this is done is by modifying the SCSI controller used to a different type. In edit virtual machine settings, if you click on the SCSI controller, you can change it so that the SCSI devices can be shared. The only problem is this change is globally on the SCSI controller, so to make this not problematic for the other SCSI devices, such as the windows C: drive, I moved the VTL SCSI devices over to a different SCSI id (1:0 instead of 0:3). This will make vmware automatically add a new SCSI controller to the box. This one I made the change and everything was happy.
The next thing I bumped into was our firewall. For whatever reason our checkpoint vpn tunnels are not setup correctly. I could not figure out how to allow the Seattle DMZ network to talk to the Tampa DMZ network over tcp/3389 (RDP). After spending some time trying to figure it out, I used a work around. I had 4 consecutive external IP's available, so I simply NAT'd these boxes out to the edge, allowed our Seattle office's external IP to RDP in on those external IP's and had it work that way instead of figuring out the VPN tunnel madness.
The last headache to figure out was the driver for the tape devices. The funny part is this is a chicken and egg problem. In order to get the driver for the tape drives installed, I need netbackup installed (it uses a symantec halfinch.sys driver). Well, during the install for the configuration portion, it wants the tape devices to be installed to configure them. In order to remedy this, Symantec added a pause mechanism to their setup scripts so the user can pause the install after the drivers are present on the system, install the tape devices from device manager, then continue the install.
On the day of the class everything ran smoothly. The only thing I forgot to do was to populate the host files of each system with the IP information on the other machines in the 'classroom'. This wasn't a big deal, the instructor was able to figure it out and work through it.
Class ran, students seemed to be happy from what I understand, and now I have a template for this build as well as the procedure (this is why I blog this stuff!) to build it back up again.