We were some of the earliest of ‘adopters’ for a paid POC to get the Cisco UCS system in-house.

Cisco sent the hardware & a group of techs to get this thing off the ground. Any new hardware being deployed shouldn’t require a group of people from the vendor to come out to set it up, should it? In today’s society, it should be user friendly, plug-n-play, which we all know Cisco to be, right? Heh, okay, or not, but anyway, our trouble had just begun.

The first day consisted of us crowded around a rack, oohing and ahhing over the new hardware, which has a striking resemblance to SUN hardware. What kind of sucked about it was the Cisco guys got in close, plugged their laptops in and set it up, leaving us, the day-to-day guys who actually would configure any new UCS systems to have zero clue what needs to be done to get it going, much less on the network.

Initial thoughts: The UI is hard to get used to and navigate, sluggish at best. You have to create a service profile before you can do ANYTHING with the blades. Creating a service profile is tricky, and involves setting up MAC pools, UUID pools, WWN pools, WWP pools, and various other pools. Then, once created, you have to assign it to a blade, then that blade boots up, runs through the Cisco boot/config OS, and sets up everything the profile tells it to. Now, finally, after all your extensive work, you’re ready to begin :) Mind you, you can create a profile and call it ‘gold’ then use it to create a template, from which you can deploy future profiles, and in turn, keep those profiles linked to the template so one change to the template propagates down to the profiles, but this can prove problematic. For instance, almost every change to the profile requires a reboot of the blade to go into the config OS, then reboot back to whatever’s loaded. Some things are intuitive, like changing the number of NICs or HBAs, but changing something simple might reboot it as well. If you change the template, it may [or may not] force a reboot of ALL linked profiles. Fun, huh??

After it’s on the network, it took Cisco something like 8 hours to configure a live ethernet port so we could fire up a blade. We had all the key players in the room working on it, but it still took them a full working day before we could even try to provision an OS on a blade. I would have bet my paycheck against any schmo who said Cisco couldn’t provision a live port in less than 4 hours, but what do I know? The on-site techs’ excuse was, ‘oh, this is the new firmware, we’ve never seen it before.’ ORLY!?!?!?!? It’s so new that it doesn’t look anything like what you’ve configured before for this hardware?

Day 1 was wasted setting up the hardware, day 2 was wasted configuring an ethernet port, day 3 consisted of trying to load different OS platforms on it (RHEL, Windows 2003 & 2008, both 32-bit and 64-bit), and although we were successful there, it took another day to get the RHEL installation flying because of the kernel version of our internally supported builds. After testing a new kernel, the network issues we had quickly vanished.

It seemed to handle every OS we support [internally] without any real issues.

Later I’ll cover what came after the long & drawn out 60+hr week of getting this far.

To come: Part 2 – my vSphere ESXi & UCS woes