Rpi3 crashes

[Laser_Man]

Hey buddy, keep calm!
Regarding the logs, I’m working on this.

IMHO, it will not be very helpful to flood the forum with a number of log entries without relation to the context where the crash happened, right? Thus, demanding for log files only sounds easy to be done on the 1st degree. But this is not true! Since OSMC does not have syslog or journald entries, there is only kodi.log which is not enough in my eyes. Who knows, if the last line in the log is the last line written? If some of the “it’s the PSU” trolls are right, nobody can guarantee, the last couple of lines are even in the logfile.

BTW:
There is no evidence yet, that this is “unrelated” as you repeatedly tell us. I always get the “lost of skin settings” after the crashes I reported. So there in fact is a relation! You might have another point of view if this is a strong relation or not. That’s okay. That’s the reason I asked for getting my parts moved into a new topic.

If this is impossible I kindly ask for getting told that.

BR,
Laser-Man

[Laser_Man]

@dillthedog

If you want to follow-up on your hunch, why not create a cron job that will look for zombie processes (not so sure about threads). Something along the lines of ps axo pid=,stat= | awk ‘$2~/^Z/ { print }’ will give you the PID of any zombie process. Just save it to a file for later analysis.

Similarly, if you think that memory is disappearing over time, the same cron job can run the free command or vmstat If things crash, you’ll be able to validate your hypothesis.

Thank you for the idea!
I just did it this way:
created script /root/cron_dump_zombies.sh

#/bin/sh
logger -i --server 192.168.0.138 "scan for zombies on $(hostname)"
ps axo pid=,stat= | awk '$2~/^Z/ { print }' | logger -i --server 192.168.0.138
logger -i --server 192.168.0.138 "scan done."

and /root/cron_dump_vmstat.sh

#/bin/sh
logger -i --server 192.168.0.138 "dump mem stat on $(hostname)"
vmstat | logger -i --server 192.168.0.138

and put both into cron (which need to be installed 1st)

    */5 * * * * /bin/sh /root/cron_dump_zombies.sh
    */10 * * * * /bin/sh /root/cron_dump_vmstat.sh

Now my servers rsyslog will get this all.
I also forward the all the KODIs logs now this way:

    tail -n 2048 -f $HOME/.kodi/temp/kodi.log | logger -i --server 192.168.0.138 &

from all boxes to the same server. the server has a rsyslog filter rule to dispatch the incoming into different files. We’ll see the outcome soon…

Any idea regarding forwarding the kodi.log early before it’s written to a physical file?
I never used named pipes, but I could imagine, this is more reliable to get the last log line when a power dropout occurs. Would highly appreciate some hints…

Finally, can I get the power state information from Raspi somehow? The same that generates the yellow flash or the rainbow square if power is under the limits. Ideas?

Cheers,
Laser-Man

[ActionA]

@Laser_Man

Since OSMC does not have syslog or journald entries,

You are most certainly incorrect here. Providing logs as directed in our wiki absolutely provides journalctl output.

@Laser_Man

IMHO, it will not be very helpful to flood the forum with a number of log entries without relation to the context where the crash happened, right?

Not necessarily… It could be possible that we see a glaring mis-configuration of your system that would easily explain your symptoms.

@Laser_Man

There is no evidence yet, that this is “unrelated” as you repeatedly tell us. I always get the “lost of skin settings” after the crashes I reported. So there in fact is a relation!

There is no relation. You are conflating two different issues and are assuming them to be the same. System crashes are a completely different issue to the skin changes being lost issue. The skin changes lost issue is addressed (previously in this thread as well as numerous other threads on the forum) as being solved and the solution is available for testing until our team is satisfied that it is appropriate and should pushed as an update to all users/platforms. I stand behind my direction provided in my last post.

[sam_nazarko]

@Laser_Man

IMHO, it will not be very helpful to flood the forum with a number of log entries without relation to the context where the crash happened, right?

Do not abbreviate logs; it’s better to have everything rather than what the user perceives to be relevant.

@Laser_Man

If some of the “it’s the PSU” trolls are right

Seriously?

@Laser_Man

The RPi3 consumes only up to ~300 mA,

Now that’s trolling.

As you seem to have it all figured out, I’m going to close this thread. If you can’t post logs, we can’t help with issues. People are trying to help you: but they’re not forced to.