service life (part 1)

disconnected (from the internet)

on the second of january, my GS724AT failed.
this switch connected all of the 10/100 or lower priority gigabit stuff offline, which included i'm generally not one to get rid of stuff if it breaks, and thus began my quest to save my son.


triaging the switch yielded a few observations: the power light was green, the data plane went thru reset as all the port LEDs flashed when reset was held, and the power supply wasn't dead as all of this happening. my theory here was that the flash had a block decay (similar to an issue i experienced with the BMC on my talos II that made it useless until i hopped thru serial to write eeprom blocks manually) somewhere in the OS payload and that was causing the switch OS to not come up. the next logical step was to get a UART/serial console on the machine while it was in its bootloader and TFTP on a new os image, as someone had done with their GS724T. this would be a bit optimistic. i consulted the datasheet of the soc on board and managed to use my multimeter to map out the pins to the JA1 header on board.

|T X X X X|
|X R G X X|

the problem? no matter what device i tried to access this uart with (teensy, FT232, raspberry pi named "uartsimpson"), i didn't get any data. my multimeter read ~4.5v to GND, and a glance on my lecroy 9324 showed it was outputting some sort of signal, but i didn't particularly trust either. thus began oscilloscope quest.

problem tree

what happens when you don't trust your test equipment? how do you solve a problem when you can't trust your measurement tools. my tektronix 2235a caught on fire 8 months ago(rifas), and my lecroy 9324 barely works and has autism. i was pretty tortured by this and decided to buy a new agilent 54622A, which of course, doesn't work. unlike some of my friends suspect, i was sold this as a working unit. butit shows no waveform trace. believing this to be related to the dead lithium battery, i opened it up to get a better look, and managed to immediately fuck up all the clips for the front panel (rip $150). even worse, i saw in the serial console that the time interpolator had failed which is why the trace wasn't showing. without schematics, i was at an impasse.

it hasn't gotten better actually

yep! this happened this morning. with the agilent scope i did buy a GSM7224 to use as a replacement, and that works! and are back up, but i've made no progres on scopequest, or really any on switchquest. in part 2 i will hopefully have actual information on how to fix these things, fingers crossed.

that's not it but i can't think of other stuff (i copied this header lol)

i had like two panic attacks in the middle of the night last night, which had me fucked up bad this morning, and i've been having a bad day but this track made it a little better when i was driving around doing random bullshit. the end. until later. whatever

published by coldmittens
last edited on 1/5/2024 at 10:47pm