0

I'm currently administrating an IBM Power 740 server via an HMC. We have four LPARs running, two VIOS LPARs, one application LPAR and one database LPAR. We have implemented lpar2rrd to monitor CPU usage and memory on the entire server (CEC), we can also monitor CPU usage per LPAR. However we can't monitor memory per LPAR.

We've been asked to graph the memory usage per LPAR so the help desk people could monitor memory in a LED display.

I know the vmstat and svmon commands; I was thinking to parse some attributes of one of these commands and graph them with Processing.

Is there any other method you know to do this?

Mat
  • 1,536
  • 1
  • 17
  • 21

1 Answers1

1

You'll have to figure out the actual implementation of getting the data out of vmstat and svmon, but once you've got the memory usage as an integer (or string, I suppose).. Sign up for an account on https://www.hostedgraphite.com/ (just the Free plan will suffice, I suspect). Then you can poke the data at them over a simple TCP interface.. 

echo "YOUR-API-KEY.metric_name 1.2" | nc carbon.hostedgraphite.com 2003

You'll need to replace YOUR-API-KEY with the value of your API key, and metric_name with whatever you want to call that data point.

I suspect netcat is installed on AIX.

I don't work for HostedGraphite, but their solution is a really neat way to get SaaS graphing for cheap.

If you've got a spare linux box around, you could also use Munin, and write some plugins to parse the output of vmstat and svmon and squirt them down a tcp socket to your munin server on port 4949. There's some Perl CPAN modules for handling Munin output neatly somewhere, and I'm sure Perl is available on AIX too.

Tom O'Connor
  • 27,480
  • 10
  • 73
  • 148