The registrations keys for the vxfencing coordinator disks are created when the vxfen driver starts up.
The vxfen driver or VCS components do not check for the registration keys periodically until vxfencing driver receive a RECONFIG message on either the primary port or the client ports which lists a node leaving the visible membership.
If lower node is remaining in the cluster, it races to grab the coordinator disks. This is done by preempting the registration keys for the lowest node that left the cluster. The node that races RACER will commit suicide(panic) if it is unable to grab a majority of the coordinator disks. If the RACER panics then all that nodes that formed cluster with RACER will also panic.
Sometimes it happens that the vxfencing registration keys for the coordinator disks may get lost due hardware issues or user error. End users can implement a cron job to check for the registration keys in the vxfencing coordinator disks by running a script once every day or couple of times a day.
-----check_registrationkeys.sh-------
#!/usr/bin/sh
Dt=`date'+%m/%d/%y %H:%M:%S:'`
[ ! -f/etc/vxfentab ] & {
echo "$DtWarning: Missing /etc/vxfentab file."
exit1
}for Dsk in`cat /etc/vxfentab`
do
echo "$DtChecking $Dsk for Registration keys ... \c"
Keys=`/sbin/vxfenadm-g $Dsk| grep "Total Number Of Keys" | cut -d: -f2`
echo " found$Keys"
if [ "$Keys"-eq "0" ]
then
echo "$DtWarning: Registration keys missing for $Dsk. Check if device is accessible andrun /sbin/vxfenadm -g $Dsk. If device went offline but came online now and keysare still there this message can be ignored."
fi
done
----------------------------------------
The output of the script can be emailed out to list of recipients
check_registrationkeys.sh | mailx -s "check vxfen keys" "$recipients"
NOTE: Vxfencemonitor is added by default in new versions starting 5.1