#!/bin/sh
# ctdb event script for monitoring the multipath daemon
#
# Configure monitporing of multipath devices by listing the device serials
# in /etc/ctdb/multipathd :
#   CTDB_MONITOR_MPDEVICES="device1 device2 ..."
#

[ -n "$CTDB_BASE" ] || \
    CTDB_BASE=$(d=$(dirname "$0") ; cd -P "$d" ; dirname "$PWD")

. "${CTDB_BASE}/functions"

# service_name is used by various functions
# shellcheck disable=SC2034
service_name="multipathd"

loadconfig

[ -n "$CTDB_MONITOR_MPDEVICES" ] || exit 0

service_state_dir=$(ctdb_setup_service_state_dir) || exit $?

multipath_fail="${service_state_dir}/fail"

multipathd_check_background()
{
    for _device in $CTDB_MONITOR_MPDEVICES; do
	# Check multipath knows about the device
	_out=$(multipath -ll "$_device")
	if [ -z "$_out" ] ; then
	    echo "ERROR: device \"${_device}\" not known to multipathd" \
		 >"$multipath_fail"
	    exit 1
	fi

	# Check for at least 1 active path
	if ! echo "$_out" | grep 'prio=.* status=active' >/dev/null 2>&1 ; then
	    echo "ERROR: multipath device \"${_device}\" has no active paths" \
		 >"$multipath_fail"
	    exit 1
	fi
    done
    exit 0
}

multipathd_check()
{
    # Run the actual check in the background since the call to
    # multipath may block
    multipathd_check_background </dev/null >/dev/null 2>&1 &
    _pid="$!"
    _timeleft=10

    while [ $_timeleft -gt 0 ]; do
	_timeleft=$((_timeleft - 1))

	# see if the process still exists
	kill -0 $_pid >/dev/null 2>&1 || {
	    if wait $_pid ; then
		return 0
	    else
		cat "$multipath_fail"
		rm -f "$multipath_fail"
		return 1
	    fi
	}
	sleep 1
    done

    echo "ERROR: callout to multipath checks hung"
    # If hung then this probably won't work, but worth trying...
    kill -9 $_pid >/dev/null 2>&1
    return 1
}

case "$1" in
monitor)
	multipathd_check || die "multipath monitoring failed"
	;;
esac

exit 0