localhost Current Load with multiple CPUs

[Box293]

I was looking at our Nagios XI server and noticed that it has a load of about 7 - 14 on average. This is a three CPU VM.

I was playing with the check_load command and noticed that there was the -r option that Divide the load averages by the number of CPUs.

without -r

Code: Select all

./check_load -w 5.0,4.0,3.0 -c 10.0,6.0,4.0
CRITICAL - load average: 14.27, 8.65, 8.25|load1=14.270;5.000;10.000;0; load5=8.650;4.000;6.000;0; load15=8.250;3.000;4.000;0;

with -r

Code: Select all

./check_load -w 5.0,4.0,3.0 -c 10.0,6.0,4.0 -r
OK - load average: 4.76, 2.88, 2.75|load1=4.757;5.000;10.000;0; load5=2.883;4.000;6.000;0; load15=2.750;3.000;4.000;0;

I then modified the Currnet Load service to use the -r. This is great, the localhost is no longer going into warning and critical states. I can also see the before and after affects when looking at the performance graphs for Current Load.

current load graphs - 4 hour.png

current load graphs - 1 week.png

However I've noticed that in the dashlet Server Statistics the load is still very high, it does not reflect the data for the Current Load service.

Question: Is there a way to make the dashlet Server Statistics match the service (use the -r option)
Suggestion: Change the Current Load service to use the -r option as the default setting. I know that the VM is distributed as a single CPU VM but it's less hassle to change it later on, also physical boxes are usually going to have > 1 CPU.

server statistics.png

Nagios XI 2009R1.4B running on a VMware VM on an ESXi 4.1 host.

[mguthrie]

I see what you mean, and this is a decent idea, although the code for this change is buried somewhere that I was unable to find at the moment. I would suggest filing it as a feature request on our tracker. tracker.nagios.com.

[tonyyarusso]

Suggestion: Change the Current Load service to use the -r option as the default setting.

Makes sense to me. I filed it at http://tracker.nagios.com/view.php?id=126.

Question: Is there a way to make the dashlet Server Statistics match the service (use the -r option)

"Yes". I found the line of code that does this, and in the current SVN copy it's #185 in nagiosxi/basedir/cron/sysstat.php, which looks like this:

Code: Select all

	// GET LOAD INFO
	$cmdline=sprintf("/usr/bin/uptime | sed s/,//g | awk -F'average: ' '{  print $2 }'");

As you can see, this particular piece isn't even using the Nagios plugin, but just the uptime command directly. I would imagine though that you could change it to do so, with something like this:

Code: Select all

	// GET LOAD INFO
	$cmdline=sprintf("/usr/local/nagios/libexec/check_load -r -w 100,100,100 -c 200,200,200 | sed s/,//g | awk -F'average: ' '{  print $2 }' | sed 's/|.*//'");

[Box293]

Cheers,
I'll have a play and see how it goes.

[rdedon]

Very nice, thank you

[Box293]

Since I made the change that Tony suggested the Server Statistics dashlet has worked fine and is accurately reporting the data.

[mguthrie]

Nice!! Glad it's working for you!