A recommended practice for monitoring DataPower is to set up a dedicated log target and file for the throttler logs. The throttler log records the memory and a few other key statistics once every 10 seconds. This data can be invaluable to understanding your systems and can also help support if there is a issue on the device. A sample log looks like this:

Mon Oct 28 2013 15:47:39 [slm][notice] throttle(Throttler): tid(1103): Memory(2934698/4148536kB 70.740570 free) Pool(820142) Ports(31799/31850) Temporary-FS(36/242MB 14.876033 free) File(OK)
Mon Oct 28 2013 15:47:42 [system][debug] : tid(127): cpu usage: 38%(10 sec) 12%(1 min) 10%(10 min) 9%(1 hour) 8%(1 day)

The most important parts of these 2 messages is the free memory percentage and the CPU averaged over the last 10 seconds. Having the history of these 2 numbers can help manage the status of the devices.

 

Configuring this is simple; first in the throttler turn on the status log and set the log-level to notice. Then define a dedicated log target. This is better than using a shared log so as to avoid losing any of the messages if the log were to get flooded by other traffic. This is how the log target should look in the CLI:

logging target: throttle [up]
------------------------
admin-state enabled
type file
format text
timestamp syslog
size 5000 kilobytes
local-file logtemp:memory
archive-mode rotate
rotate 10
feedback-detection off
event-detection off
suppression-period 10 seconds
event system debug
event slm debug

The size can be adjusted and this can be useful if, for example, you have an older box without RAID. Since the messages only use about 500 bytes every ten seconds you could keep a lot of history in just a few megabytes of files (~5MB per day).

These files can also be easily parsed to generate graphs of the CPU and Memory.

 

Tags:  logging datapower throttler throttle monitoring memory websphere log