HP-UX Workload Manager User's Guide

Advanced WLM usage: Using performance metrics
Supplying data to WLM
Appendix H510
Use wlmsend on the command line, in a shell script, or in a perl program.
Use wlmrcvdc in the WLM configuration file.
NOTE Avoid the process overhead of using wlmsend in a compiled data collector;
let wlmd invoke your data collector through the configuration file if
possible. Use the API described in “Sending data from a collector written
in C” on page 499 to write such a data collector.
The wlmrcvdc utility creates a rendezvous point and forwards data from
that rendezvous point to WLM. Use wlmsend to send data to a
rendezvous point to be collected by a wlmrcvdc invocation.
If you use wlmsend, you must also use wlmrcvdc. You can, however, use
wlmrcvdc without wlmsend if you specify wlmrcvdc with a data-collecting
command.
The syntaxes for the commands follow:
wlmrcvdc [-h] [-V] [-u user] [-g group] [command [args...]]
wlmsend [-h] [-V] [-w wait_time] metric [value]
For an explanation of the syntaxes, see Appendix A, “WLM command
reference,” on page 363.
NOTE Be careful of I/O buffering when feeding data to wlmsend. For example,
the following line works as expected:
tail -f logfile | /opt/wlm/bin/wlmsend job_time
However, adding awk or other utilities that buffer I/O can result in data
being delayed, or not even sent, to wlmsend—depending on when the
buffer is flushed:
tail -f logfile | awk ‘{print $1}’ | /opt/wlm/bin/wlmsend
job_time
Always check the file /var/opt/wlm/wlmdstats, created by wlmd -l
metric as explained in wlmd(1M), to ensure that your data gets to WLM
in a timely fashion.