Server randomly crashes/is inaccessible but the bigger problem is my physical power button doesn't do clean shutdowns

[s449]

About 3 times in the last year I notice my server is randomly inaccessible. Plex is down, the webGUI is inaccessible, and my SuperMicro IPMI interface is inaccessible. I think it usually happens overnight.

 

I finally got a spare monitor so this time I plugged it into my machine to debug and saw just a black screen. I assume it's related to this:

 

I do have an Nvidia 1060 GPU passed through to Plex for hardware encoding/decoding.

 

But debugging all of the above issues aside, the bigger issue is when I push the physical power button on my Node 804 it does nothing. Usually when I initiate shutdown via the webGUI, immediately I hear some beeps and it goes down. But I've waited 10+ minutes after hitting the physical button, no beeps, no shut down. I've pushed it a number of times, it doesn't seem to work.

 

However, if I hold it for around 5 seconds, it does an unclean shutdown (like the power cord was ripped out). So it does seem to be connected somehow. But maybe not connected the right way?

 

Maybe this is simply a hardware issue or maybe something is making it stuck like on Windows "[software] is preventing shutdown". Any ideas? Thanks!

[JorgeB]

On 7/16/2022 at 7:59 PM, s449 said:

But debugging all of the above issues aside, the bigger issue is when I push the physical power button on my Node 804 it does nothing.

Do you mean every time or when the server is inaccessible? If just when the server is inaccessible it's normal if the server crashed.