#!/bin/sh
# script to check accessibility to the reclock file on a node

. $CTDB_BASE/functions
loadconfig ctdb

cmd="$1"
shift

PATH=/usr/bin:/bin:/usr/sbin:/sbin:$PATH

# The size of this file represents the number of intervals that have
# passed when we have tried to but failed to stat the reclock file.
# after third failure the node becomes unhealthy
# after the twenteth failure the node we shutdown ctdbd
RECLOCKCOUNT="$CTDB_BASE/state/reclock-fail-count"

case $cmd in 
     startup)
	echo -n > $RECLOCKCOUNT
	;;

      monitor)
	echo -n 1 >> $RECLOCKCOUNT

	COUNT=`ls -ln $RECLOCKCOUNT | cut -d" " -f5`
	[ $COUNT -gt 19 ] && {
		echo "Reclock file can not be accessed. Shutting down."
		sleep 1
		ctdb shutdown
	}

	RECLOCKFILE=`ctdb -Y getreclock`
	[ -z $RECLOCKFILE ] && {
		# we are not using a reclock file
		echo -n > $RECLOCKCOUNT
		exit 0
	}

	# try stat the reclock file as a background process
	# so that we dont block in case the cluster filesystem is unavailable
	(
		stat $RECLOCKFILE
		[ "$?" -eq 0 ] && {
			# we could stat the file, reset the counter
			echo -n > $RECLOCKCOUNT
		}
	) >/dev/null 2>/dev/null &


	[ $COUNT -gt 2 ] && {
		echo "Reclock file can not be accessed. Mark node UNHEALTHY."
		exit 1;
	}
	;;
esac

exit 0