I learned a couple of lessons about the physical side of server management these past few days. The first lesson was to test the Ethernet cables before installing them. The second was that the server put out a gigantic amount of heat. Both caused me some painful and unexpected learning opportunities.
I had purchased a set of CAT6 cables on Amazon. It hadn’t occurred to me that some might be bad—but it turns out they were. I was checking the status of the ports on my Dell switch using
show interfaces status and saw that some of the ports were running at 100MB instead of 1000MB. I though it was a configuration problem and spent hours tweaking parameters such as
flow control but all to no effect.
In desperation I decided to run a cable diagnostic. Bingo, that was it! The test cam back with “open cable” error and a “cable short” errors on the problem ports. The diagnostic was easy to run:
test copper-port tdr g[port #]. I swapped the cables and re-ran the diagnostic. All good! From now on I will test every cable before installing it. I promptly ordered an Ethernet cable tester.
The second lesson I learned was that these servers generate a lot of heat. This is obvious in hindsight but I hadn’t anticipated it. The room were I keep the rack is not large which compounds the problem. They ran so hot I had to shut half of them down before they all failed. I decided I would need additional cooling so I ordered a portable AC unit. Hopefully, this will solve the problem. I will know soon once the unit arrives.
Update: The portable AC unit worked. I did have to remove the filter from the XRackPro case however. The filter simply didn’t allow enough air to flow to keep the servers full. I think I will buy some nicer case fans as well. The Noctua fans look like they can push more air more quietly than the stock fans.