NET/MASTER Management Services (MS) System Management Guide

Managing the Activity Log Message Queue
Managing Activity Logs
10–6 105744 Tandem Computers Incorporated
Refer to the NonStop NET/MASTER NCL Programmer’s Guide for an explanation of how
to write and modify a LOGPROC procedure.
Specifying the Active
LOGPROC Procedure
Use the LOGPROC operand of the SYSPARMS command to specify the name of the
active LOGPROC procedure. The command takes this form:
SYSPARMS LOGPROC= logproc-name
Specify logproc-name as the name of the NCL LOGPROC procedure. The
procedure is normally stored in the customized NCL procedure library.
Deactivating the Active
LOGPROC Procedure
To terminate the operative LOGPROC procedure, issue this command:
SYSPARMS LOGPROC=FLUSH
After the LOGPROC procedure is terminated, all messages proceed directly to the
activity log.
Managing the Activity
Log Message Queue
Messages destined for the activity log arrive sequentially in the activity log message
queue. When a message arrives on the queue, it is logged immediately if there are no
other messages on the queue: otherwise, it is queued for logging. When a message is
logged, it is written to the activity log on disk as rapidly as possible.
NonStop NET/MASTER MS can read a message and place that message on the queue
faster than it can write a message to the activity log. Therefore, if NonStop
NET/MASTER MS reads many messages over an extended period of time, the size of
the queue increases faster than messages are logged.
If the size of the queue continues to grow, NonStop NET/MASTER MS may
experience severe performance degradation due to increased memory consumption.
Two common situations may cause the size of the queue to continue to grow. For each
situation, you should take the corrective action, described below, to ensure that you do
not experience performance problems.
First, a local NCL process may generate an unexpected number of messages reporting
an event or events, and send the messages to the queue more rapidly than the
messages can be logged. The solution to this problem is to monitor all NCL activity
carefully, to flush an NCL process that generates an unacceptable number of messages,
and to rewrite the NCL procedure so that it does not generate an unacceptable number
of messages.