BASH/CRONTAB - Restart ethminer service if stalled

ethminer-monitor.sh

#!/bin/bash

PATH=$PATH:/usr/bin:/usr/sbin:/var/log/miners:/home/ubuntu/scripts
SECS=$(date +%s)
ETHMINERLOGFILE=/var/log/miners/ethminer.log
ETHMINERRESTART=/var/log/miners/ethminer-restarts.log
idletime=$(expr $SECS - $(date +%s -r $ETHMINERLOGFILE))

echo -n "$(date) - Status check running.. " &>> $ETHMINERRESTART
if [ "$idletime" -gt 300 ]
then
        echo "$(date) - ETHMINER process has been idle for: $idletime seconds" &>> $ETHMINERRESTART
        echo "$(date) - Restarting..." &>> $ETHMINERRESTART
        echo "Last 20 lines of old log:" &>> $ETHMINERRESTART
        echo "" &>> $ETHMINERRESTART
        /usr/bin/tail -n 20 $ETHMINERLOGFILE >> $ETHMINERRESTART
        service ethminer restart
else
        echo "OK" &>> $ETHMINERRESTART
fi

@ddobreff - Yeah, that was something I mentioned in comments on Github (where I posted these gists). But for quick/dirty way - the log file modified time works well for me. Another problem is the binary tends to just 'stop' (and not die/become defunct/etc) when it does 'go down'. So 'watching' the console output AKA log file is a better method in this case. Till he 'defunct's' or exits process in source code on a network timeout(?) (my guess as to the halt condition), the PID method just wont work. It will continue to think binary is running - because it is.

Just updated it again - if using log area defined (/var/log/miners/*), you must first sudo mkdir /var/log/miners and then own it to user running daemon. Example: sudo chown -R ubuntu:adm /var/log/miners if running service as user 'ubuntu'.