top of page

Cisco 5108 Blade Server Chassis received!

The box has arrived!

The gentleman from the freight company wheeled this guy right into my garage with his pallet jack and set it down. After unwrapping the polyurethane (trash bags?) from the outside, I have to laugh at the box used to cover the bubble wrap! There had better not be 300 lbs. of dog food in here!

It is not recommended that you try to pick this thing up by the exterior handles to move it. As I mentioned in my previous post, this thing is almost 300 lbs fully loaded, so you have to pull all of the blades, fans and power supplies out before moving it. Once you've done that, the remaining 100 lbs isn't too difficult to hoist up on the workbench.

You can see why these things require four post racks. Apart from the weight, they are extremely long and just wouldn't be properly supported by only screwing in the front rack plates...and since my rack rail is bent, I think that's exactly what someone tried to do in the past. This is no Raspberry Pi lab. Go big or go home!

One of the things you'll notice when you go into the Cisco Power Calculation Tool is that there are two versions of the blade chassis, both with the same model number (5108). I've definitely got the V1 model, and after extensive research from the original spec sheet, these gold supplies (N20-PAC5-2500W) are rated for voltages as low as 100V, even though the sticker on the top indicates a 200V minimum. It turns out, the spec sheet is wrong! (not the first time we've seen that, right?

After much more googling and reviewing a hundred links, I finally find the following post on the Cisco Community page about plugging 110V into my server. True to the information in the post, plugging 110V into my power supplies does exactly nothing. No LED's are lit on the front of them, which means they are not getting power. The post does indicate a work around, though.

There are 3 options at this point.

1) Run 208V circuits to the garage

2) Buy a pair of transformers

3) Replace a pair of the power supplies with the ones from the UCS Mini Chassis (UCSB-PSU-2500ACDV)

I've decided to with option 3. Here's why:

Option 1: I can't run a 208V circuit from my breaker panel without redoing a bunch of walls because the the panel in the basement and the basement was finished. I could cut a hole in one wall of the garage and find the 208V circuit that my dryer is on and tap into that (the laundry room is just on the other side of the interior garage wall, probably like a lot of houses), but I don't really know if the dryer and blade server will happily run at the same time on that circuit...and our dryer is in use all the time (far more than the microwave, even) so this is not a great option.

Option 2: A high quality transformer runs anywhere from $75 to $125. It's not a cheap solution, and I will lose some efficiency in the process (i.e. increase current draw), and lose some percentage of the effective breakered amperage in my circuit. Plus, I am still using the old power supplies, which means I can't sell them to help pay for the new ones.

Option 3: Buy the pair of replacement power supplies for the UCS Mini Chassis, which go for around $40 on Ebay, shipping included. I can also sell the old ones for about $40 (shipping included), so if I buy two (more than enough to support my Max 1400W installation) and sell four, I should come out ahead on the deal.

Continuing our inspection, we look at the rear of the chassis and see that we got 2204XP fabric extender cards instead of 2208XP's, the difference being 4 vs. 8 10G ethernet ports. I'm not too bothered by this error, as I think a single 10G interface will be all I'll ever need for my #homelab, but 8 is definitely enough.

The new power supplies are ordered and should be here by next Thursday. With any luck, I'll have a video to post next Friday with a walk-through of the chassis with a successful power-on.

99 views0 comments


bottom of page