TRANSFER Installation and Management Guide

Sample Monitor Configuration
TMANAGER
13198 Tandem Computers Incorporated 10–25
Considerations in Configuring TMANAGER
Consider the following as you configure TMANAGER monitoring:
TSAMPLE counts the records in the READY, TIME, and NET queues based on the
queue sample time interval. If the time interval is set to a low number (such as 15
minutes) and the queues have a large number of records (such as 20,000), the
TSAMPLE server constantly reads the queues and can contend with TRANSFER
operations.
Because the TIME file is potentially large, sampling of it is disabled by default. If
you wish to monitor the TIME file you must enable sampling. See “Configuring
TMANAGER” later in this section.
Active alarm notification is done on a nowait basis; if it cannot be completed in
five seconds, the alarm message is canceled and TSAMPLE continues processing.
The active alarm notification is repeated at the next sample interval if the alarm
condition still exists.
Because held records are included in the counts of records in the READY queue,
specifying a threshold for the READY queue includes both held and unheld
records.
The appropriate queue thresholds and types of work to be monitored vary from
system to system. In a small TRANSFER system, 10 Submit Pkg requests in the
READY queue might be abnormal, but a large system might routinely have 20 or
30. Similarly, a system that is part of a multinode network should monitor the
NET queue to keep abreast of potential network problems, but it is not necessary
for a stand-alone system. As a starting point, use Scheduler Queue Samples to
approximate normal queue thresholds for the types of work you want to monitor.
Sample Monitor Configuration
A sample configuration to help you monitor your TRANSFER system might look like
this:
Active Alarm Notification: Y
Active Alarm Log Location: \MYSYS.$0
The active alarm alerts you to potential problems in the network. Select a
prominent log location so that you can see the alarm when it is activated.
Database Sample Interval: 4 H
This value depends to some extent on the value of the database threshold. The
higher the database threshold, the less time you have to react to the database’s
becoming full. Because the system overhead in monitoring the database is
minimal, there is no real penalty for increasing the frequency of the sample
interval.
Queue Sample Interval: 1 H
In order to monitor problems with the asynchronous part of TRANSFER, you
probably want to sample the queues every hour.