Debugging a system crash that produces no error

Got a weird one here...

I've got a machine running 13.1 where if I try to start logstash or graylog, the machine locks up entirely within a few seconds - no response to pings or to typing at the console, and the fan spins up somewhat. This happens whether I run the service on bare metal or in a jail. No error is reported anywhere: the console doesn't show anything beyond the last command and where java was found; the contents of /var/log just show a reboot with no indication as to why; there's nothing in /var/crash. I've tried running the command under truss to see if it's doing anything weird, but all I see is freads and mmaps until it stops. Watched memory use with htop while it happened, but memory use was below 10%.

I've checked my disks (stress test with fio and scrub of my non-root zfs disk), and they seem to be fine. The RAM's been replaced, with the issue occurring both before and after. It's a second-hand M715q with an Ryzen 3 Pro 2200GE - suppose it could be a hardware issue, but it's only these processes that fail, with compiles and other daemons running fine.

Any thoughts on where I could look next?
 
Have you looked at /var/log/logstash/logstash-plain.log?

I have logstash 8.6.0 running without issues. Before I upgraded it, logstash 7 ran without problems too.
 
Generic kernel. I'll try disabling some more boot services, see if that changes anything. I don't have logstash starting automatically, btw, so if I don't start logstash manually then the machine runs indefinitely without problems.

/var/log/logstash doesn't exist. I don't think the process gets far enough to create it. It'll crash the system even when given just the --help flag.

Will check out mprime, thanks. I suspect it might be hardware-related because I've got another embedded machine that doesn't have any issues running logstash, but that one's also very different hardware. Also, if it is faulty hardware, why is java so good at triggering the fault? ?
 
Back
Top