AbydosOne

Alright, I'm back.

The last day of Lent was spent moping because I killed a motherboard and/or CPU by sending 12 volts to a pin it's not supposed to go on. Mostly sucked because it was pretty much the last step in a project I have to set up an external, SAS expanded drive tower, which actually pretty much all worked (much to my surprise; I'll do a write-up when I have everything dialed in and tested).

I flew too close to the sun and decided to use a TRRS connector to send the 12V signal to trigger the secondary PSU and send PWR/RESET back; it works fine if you don't hotplug it, which is a design consideration I had reviewed, but didn't look closely enough at. Hotplugging would short the PWR button, which would start the board, which would start the host PSU, which would then have 12V on the tip of the TRRS, sending 12V into the rest of the pins. Oops.

I haven't determined if it's just the motherboard, or if AM4 has some of the PWR/RESET functionality baked into the CPU, but the board won't proceed anything past initial power on. At least it's a pretty cheap AsRock board that may still be under warranty (it's not technically mechanical damage ¯\_(ツ)_/¯), and a pretty middling 1600X in it. Tomorrow's project is to pull down the server for a little bit to do a CPU swap to at least prove that the board is bad (this CPU is incompatible with the server board, so it's only a one way test).

As far as I can tell, if I hadn't put in the PWR/RESET functionality, it would work fine, so I'm going to go back to that and chalk this up as a particularly expensive EE learning experience. Now I know.

FYI: I'm going AFK for the next ceil(40+(40/7)) days (give or take) for Lent (the disparity from "40 days" is due to quirks between Protestant and Catholic Lent traditions, in case you're curious).

While LTT isn't what I would particularly consider "social media" by strict standards, my thinking is more along the lines of "if I have to ask, the answer is probably 'yes'." I've already logged out of FB on a few devices. I'm mulling over cutting YT as well...

See you all in a few weeks!

My Proxmox server hung while I was taking a nap because *checks notes* it ran out of disk space? Turns out setting a virtual disk to "backup=0" doesn't prevent it from being snapshotted by ZFS, and so that ate up ~250GB on a disk that gets a lot of scratch use.

In other news, my new Intel X550-T2 arrived, so this weekend is going to involve putting together a testbed and installing Proxmox + TrueNAS on it to see if I can replicate the network performance degradations I'm seeing on my "production" server using the parts from the previous one.

If anyone is thinking about 10G-baseT networking, please consider your own time and sanity first. I keep getting pulled into asinine troubleshooting sessions because of a) my server, b) my assortment of 10G cards, c) my Ubiquiti Flex XGs, and/or d) my existing Ethernet infrastructure (that I can't easily replace).

Current issue(s):

• 2.5G NIC in Cable Matters USB-PD adapter seems to be flakey while PD powered (was replicable, then I unplugged it for a day and messaged Cable Matters; NIC took threat of being replaced seriously and is now behaving)
• X540-T2 in server can send data at 10G, but can't accept a single connection at 10G; can accept multiple parallel connections at 10G. Issue weirdly scales to 2.5G, even though 10G tests will almost all be >2.5G, meaning I can get 50-80% of any link for any one connection. I haven't ruled out any motherboard/host OS issues...

I'm probably going to have one of every 10G NIC by the time I'm done with this process. I've had:

• 2x ConnectX-2s - one in MCPC, one dead
• 2x ConnectX-3s - one in desktop, one dead
• 1x X540-T2 - in my server, dying? (sensing a trend yet?)
• 1x X550-T2 - coming from eBay, because I'm not going to second-guess a $15 discount

I've spent way too much of my life fretting over iperf3 results...