Following on from last week's blog.
So we bought the hardware, and after it had been delivered, installed everything in the rack, and sat back to start planning the installation. I started up one of the host machines to get a look at the POST and boot processes. To my surprise, an operating system had already been installed - and it was Windows Server 2008 R2 Datacenter. We had purchased the licences for this, but hadn't expected that they would pre-install it.
Well no problem, just have to install the VMWare ESXi. I had a version of the ESXi software, but it was an older version, so first I had to download an updated version of the software which was an .iso image, then create an install disk. Having created the disk, I was then able to do the install. I was really quite surprised; it went through very quickly. Very little to see, just a few linux type screens showing the progress of the install. But after just under 15 minutes, it was all done.
So obviously, it also made sense to do the other two hosts at the same time. Away I went and the second machine was done in much the same time, everything complete with no issues. I then started the third machine, and decided to go for a quick cup of tea as there seemd to be no point in me hanging around watching a series of dots advancing across the screen.
But when I got back, I had a bit of a shock; the process had stalled part way through. The equipment didn't seem to respond to any keystrokes, so I took the disk out to check if there was a fault, but it didn't seem so. I tried to start the install again, and unfortunately, once again it stalled. A third attempt fared no better, so I decided to take a break and look at the vSphere client install whilst I thought about what could be the issue.
I already had installed a copy of the latest version of vSphere client on my laptop for our test a short while ago, and just had to change the logon details. It connected to the host machines without any issues and I could play around with the various bits. I even did a quick install of a guest Operating System to create my first Virtual Machine. Everything looked really good.
However, I then noticed that there seemed to be something odd about the disk allocation on the datastore on the server. There were several partitions, none of which I had created. Worse, it seemed that several of these were unusable by either the VMware or by the guest OS. Having given it some thought, it seemed to me that when the ESXi software was installed, it didn't re-partition the disk in the way that might be expected, and part of the disk would never be available for use, which might be an issue.
At that point, it seemed appropriate that I should go back over the ESXi software install. I did this, checking the process, and at no point did it actually indicate that there was an option to manage the partition. In the end, I simply put the Windows disk in, then used the install routine to start up, and delete all existing partitions. After that, I ran through the ESXi install, and this time, it made all of the disk available for use. I then decided that I would do the same on the others, and the second machine completed without any issues.
The third machine also allowed me to delete the partitions OK and there seemed to be be no reason why the ESXi software shouldn't install. But still it would only go so far, then it stalled everytime. I went through this a couple of times, before going back to my desk to give it some more thought. And at that point, I discovered the reason why, and it was so frustratingly simple, I am almost embarassed to tell you what it was.
We use a very clearly structured IP address range within our network; servers get a static address in one subnet, and all addresses assigned via DHCP are in a slight different subnet. The address that I had input as part of the install routine was an address within the server range and one that had been specifically reserved for the virtualised platform.
But somehow, the address allocated for the third machine had also been given to a secondary network card on an old server. Someone had added a cable to the NIC and then plugged it into a network point. The install routine had failed because it detected that the address I tried to give it was already in use! Once I sorted out the superfluous NIC, the install routine went through without any more issues.
At this point I had 3 host machines, all installed and a connection to each tested with the vSphere client software. A good start and I felt that I was starting to understand VMware. I still had a few other things to go over, but I was feeling really quite positive about the various processes and was looking forward to getting on with it.
But the next step will have to wait for another day 8-)