A problem has been found in the Cluster Server monitor script at the 5.1, 5.1 MP1, and early 5.0 release levels where the monitor script will not catch terminated daemons that are configured to be monitored in the VCS_NBU_RSP configuration file. The problem stems from the population of the .NBU_STATE file with a missing new line at the end of the file. As a result the monitor script is unable to properly read the STATE of the cluster node and mistakenly believes that it is offline and therefore does not need to be monitored.

Verification of this problem can be seen in the /usr/openv/netbackup/bin/cluster/vcs/VCS_AGENT_DEBUG.log file as missing information between bringing the system online and offline as shown in the example excerpt listed below (note the missing state changes as daemons are dropped while the cluster node is online):

  Wed Oct  6 14:22:02 2004 Start Online.......

  Wed Oct  6 14:43:54 2004 Start Offline.......

The engineering approved workaround for this problem is to edit line 101 in the /usr/openv/netbackup/bin/cluster/vcs/monitor script which should look as follows:

  chomp($NBUSTATE=`grep "^STATE" $NbuRoot/netbackup/bin/cluster/vcs/.NBU_STATE | sed -e s/STATE=//`);

Change this line to:

  chomp($NBUSTATE=`grep "^STATE" $NbuRoot/netbackup/bin/cluster/vcs/.NBU_STATE | nawk '{print $1}'| sed -e s/STATE=//`);

The insertion of the nawk '{print $1}' piped entry appends the required new line into the expression which therefore allows for the proper population of the $NBUSTATE variable. With this variable properly instantiated, the monitor script will work as expected.