Jetson Nano stops running

I’m running a Jetson Nano headless with continuous jobs from Rosetta@home. Lately I’ve discovered that, once a day or so, the Nano has stopped running: the fan is still on, but the green power light is out and the device is unresponsive to ssh. I have to pull the power plug and reinsert to get it running.

I could guess that this is either a heat issue or a low power issue. I’ve run journalctl but I don’t see an obvious answer. Is there a log file or key phrase that I should look for?

I have no idea what the actual problem is, but the way to log this is via serial console. Serial console is immune to many failures, and won’t care if much of the system is down. If there is a log message, then it is likely something would be logged (and the serial console program runs on the PC, so any log is independent of storage on the Nano). See:

I became aware of the serial console approach when dealing with a different problem that turned out to be a corrupted SD card. I bought a USB device, wired it up, and was not able to get it to communicate anything.

Whether or not I get a serial console to work, I need to be able to parse the thousands of lines in any log file to find something that’s relevant to an abnormal shutdown. My posting was really a query to see if there is are standard or typical log messages having to do with (1) overheating; (2) low voltage; or (3) something else that’s an obvious possibility. If the Nano does not provide such log messages, that’s useful to know, but then I won’t learn anything new via the serial console. If the Nano does provide such messages, then I will look first in the journalctl and then, if I don’t find anything, try once more to get a serial console running.

sometimes unsufficient power source would cause similar issue at my nano set up;
Once the power source was replaced to a more consistent/powerfult the issue resolved.

That in fact is what I suspect, but I thought I’d check logs before purchasing a dedicated power supply for this one device (I have a half dozen SoC devices). At least one of them (either a Raspberry Pi or an Odroid N2) actually puts out a low-power log message, and I was wondering if the Nano does the same.