add an eventscript to monitor that the multipath devices are healthy

diff --git a/config/events.d/20.multipathd b/config/events.d/20.multipathd
new file mode 100644 (file)
index 0000000..93a58ab
--- /dev/null
+++ b/config/events.d/20.multipathd
@@ -0,0 +1,99 @@
+#!/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 ..."
+#
+
+PATH=/bin:/usr/bin:$PATH
+
+. $CTDB_BASE/functions
+loadconfig ctdb
+loadconfig multipathd
+
+cmd="$1"
+shift
+
+[ -z "$CTDB_MONITOR_MPDEVICES" ] && {
+       exit 0
+}
+
+MPFAILURE=$CTDB_BASE/state/multipathd/failure
+
+multipathd_check_background()
+{
+       for DEVICE in $CTDB_MONITOR_MPDEVICES; do
+               # check that we can see all devices
+               if [ -z "`multipath -ll $DEVICE`" ]; then
+                       echo Device $DEVICE not known to multipathd
+                       exit 1
+               fi
+
+               # check that all devices are active
+               if [ -z "`multipath -ll $DEVICE|grep prio|grep active`" ]; then
+                       echo Device $DEVICE has no active paths
+                       exit 1
+               fi
+       done
+       exit 0
+}
+
+multipathd_check()
+{
+       # run the actual check in the background since the call to 
+       # multipath may block.
+       (
+       multipathd_check_background &
+       pid="$!"
+       timeleft=10
+
+       while [ $timeleft -gt 0 ]; do
+               timeleft=$(($timeleft - 1))
+
+               # see if the process still exists
+               kill -0 $pid > /dev/null 2>&1 || {
+                       # it doesn't exist, grab its exit status
+                       wait $pid
+                       [ $? = 0 ] || {
+                               echo "20.multipathd: multipath background update exited with status $?"
+                               touch $MPFAILURE
+                               exit 1
+                       }
+                       rm $MPFAILURE 2>/dev/null
+                       exit 0
+               }
+               sleep 1
+       done
+       echo "20.multipathd: Callout to multipath checks hung."
+       touch $MPFAILURE
+       exit 1
+       ) &
+
+       if [ -f $MPFAILURE ]; then
+               return 1
+       else
+               return 0
+       fi
+}
+
+case $cmd in
+     startup)
+       # create a state directory to keep/track the multipath device
+       # state
+       /bin/mkdir -p $CTDB_BASE/state/multipathd
+       exit 0
+       ;;
+
+     monitor)
+       multipathd_check
+       [ "$?" == "0" ] || {
+               echo 20.multipathd: monitoring of multipathing failed
+               exit 1
+       }
+       exit 0
+       ;;
+esac
+
+# ignore unknown commands
+exit 0