[Dxspider-support] 1. Cluster started hanging several times a day. (David Spoelstra)

[Andy Scott]

I cannot offer any conclusive feedback, but I have found, on raspberry pi,
that a
symptom of failure on my Pi is that the cluster service fails to
communicate with clients on the user port. I have set a cron task to see if
a login: prompt is received when telneting into the user port (note that
you cannot rely on a telnet timeout, as communication with the port still
works fine).  If it times out waiting for the login prompt, I kill the
dxcluster process and restart the service - the cron script runs one minute
before the normal dxspider cron to reconnect to the upstream server.

In doing so, I haven't needed to reboot the server, although this may still
be needed from time to time if a different symptom presents.

The cron script is as follows, in case it helps anyone (the  /var/tmp
folder should exist and be writeable):

#!/bin/bash
#
# First, capture the process ID for the dxspider server
#
PID=`pgrep cluster.pl` ;
#
# Test that we get a login prompt - not just that the telnet port is open
# We're going to send  a login of _quit_  which will cause the cluster
# to reject the login and close the session after writing a trace to
# a temporary file that we can test later, checking its file length
#
echo "quit" | nc -w 2 [your server address] 7300 &> /var/tmp/dxtest
grep -i "login" /var/tmp/dxtest
test=$(cat < /var/tmp/dxtest)  || exit  #      exit if the file doesn't
exist
#
#  If the tested file length is long enough to include "login" then all is
working
#
if [ "${#test}" -lt 5 ]
        then
        # echo "The DXCluster has falledn over..."
        kill -9 $PID ;
        sleep 3
        # echo 'reset by checkcluster.sh' | system-cat -t DXCluster -p
warning ;
        /etc/rc.d/rc.spider
        sleep 1
        logger "DXCluster reset -p warning"
fi
exit
# The End
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.tobit.co.uk/pipermail/dxspider-support/attachments/20190307/27c1c4fd/attachment.html>