Validating load average report

[russbutton]

I've set up a Cacti server to monitor the CentOS machines in our data center. Not terribly hard. I loaded in the snmp packages through yum. Did a service snmpd start and then pointed the cacti server at 'em. Lo and behold, I get graphs! Cool.

We're trying to do some performance tuning of our app, which of course is something which works across various machines - mysql, web app, hadoop, yadda, yadda... What's of interest of course is machine load average, but I'm not sure I trust Cacti. The reason is that there's one machine in the bunch which is completely idle. Logging in on a terminal session, when I run "uptime" it shows me a load average of 0.00 0.00 0.00. It's doing *nothing*.

But Cacti gives me a load average graph floating around 1.0. As I write this, Cacti reports that the load average is 0.62 0.26 0.19, but uptime says 0.00 0.00 0.00. Somebody is lying to me. Cacti, is that you?

I tried using snmpwalk and such to see what I could get out of snmpd at a line command level, hoping to see what it is that Cacti is seeing, but no luck.

Russ

[gandalf]

The "error" is, that the values are stacked. That in fact does not make sense. But it looks nice
Change from STACK to LINE and you're safe
R.