The Great Rack Migration – D1518

The Great Rack Migration – D1518

So after a number of years labbing, I decided it’s time to graduate to a rack. This is my series of posts documenting the migration.

I deal with a lot of storage in my lab, and a huge focus has always been keeping the lab as quiet as possible. This meant I deployed everything in tower cases.

While this goal of “as quiet as possible” has been realized, what’s also been realized is a mess.

A huge mess.

This was a good day after a cleanup:

And this was the results of that cleanup:

the results on a bad day

So maybe it’s just me growing older (and my hearing going a little), and having a desire for a bit more organization, but recently I decided to bite the bullet and get a rack, the Startech 25U adjustable rack to be exact.

The original home

The first server to be migrated was a trusty old D-1518 fileserver in a U-NAS case.


I will say, that U-NAS case is pretty nice. It’s all rubberized and has a nice premium feel to it. The inside of the case is pretty tight, and I threw the fan block from the original SYS-E300-8D case to keep the CPU and HBA nice and cool:

u-nas guts

The New Home

The new home for this server was a Supermicro CSE-826 case.

so much room for activities!

The first hurdle presented itself shortly after pulling the HBA. The 2U case only takes half-height PCIe cards, and my bracket was full-height:


Fortunately, by some miracle, I managed to track down the shorter bracket in a box somewhere:

short bracket

The U-NAS looks a little sad when empty:


The Issues

No build of mine EVER goes off without a hitch. It’s the curse of being a Murphy.

Some of them, I can plan for:

  • The backplane of this server is SAS2. My HBA is SAS3. So I made sure to have a SFF-8643->SFF-8087 cable on hand.
  • The backplane of the server is an “A” backplane. This means it has internal breakouts. The short explanation here is that I need at least 3 SAS ports for all 12 bays to work, and I only have two. So I made sure to order an Intel SAS Expander when I ordered the case.
  • There wasn’t going to be room for the SAS expander to plug into the motherboard. So back to the old “powered riser” trick, which I made sure to order.

Some of them though, I did not plan for. Like the Flex-ATX motherboard being way too short for this case. That mean the 24 pin power and fan cables were all too short:

too short

I actually had a 24 pin atx extender on hand from another project, but I had to wait for a few days for some four pin fan extensions to arrive:

atx extender

The final issue I had was that the powered riser took a molex connector, and based on where I had to place the SAS expander, there was no way in hell that a molex connector would reach. So I had to order a molex extension cable as well.

Build complete

Once all the extra cables arrived, the server finally went together smoothly:

Complete server

Notice the rockstar placement of the SAS expander.

The final piece of the puzzle was that no matter what I did, the CPU ran VERY hot. So I used the old trick of placing a fan on top of the heatsink (picture was from when I did it prior).

Fan on top of CPU

Finally all racked up:

Overall, I’m pretty satisfied. It’s a little loud, but not terribly. And this is a server that’s not powered on full-time.

To be continued on the D-1521.

1 Comment

  1. dlee

    Hi. Such fun to read your networking insights and the pain going through the process of setting up new servers. You said in one of your migrations post back in 2017, that you won’t ever try to build a 1U rackmount custom build ever again. I am curious, what made you go through this pain all over again? giving secondhand prebuilt servers are cheap nowadays

Leave a Reply

Your email address will not be published. Required fields are marked *