I recently installed PMP 1.1 to monitor an Apache 2 server (cacti), and I’ve got everything running just fine, but I need some help interpreting the graphs.
The ‘Apache CPU Load’ graph shows values on the Y axis with a “m” suffix (100m, 120m, 140m, etc), but I don’t understand what that means. Is that a mean percentage over a period of time? I checked the plugin data source values and it looks like it uses an RRD “GAUGE” data source type with a 300 step, but I have no idea what that really means. The values on the graph don’t seem to correlate to the Apache web-hosted /server-stats values.
The “Apache Requests” value also doesn’t make sense to me… do these values represent average concurrent requests per time period?
The “Apache Scoreboard” is the most descriptive graph, but again some of the values (reading request, closing connection, etc) have “m” suffixes and I’m lost on what that means.
To go a step further, it would be helpful if there was something that made it clear what sort of thresholds a value should / shouldn’t have, in terms of best practices / performance. That way we can leverage other Cacti plugins like “thold” to respond to these PMP graphs if certain conditions are met.
About m:
10^-18 a - atto
10^-15 f - femto
10^-12 p - pico
10^-9 n - nano
10^-6 u - micro
10^-3 m - milli
0 (no unit)
10^3 k - kilo
10^6 M - mega
10^9 G - giga
10^12 T - tera
10^15 P - peta
Can you please attach some sample graph and post the corresponding /server-stats data?
I’ve attached my CPU Load graph which shows current value of 23.2m. At that time my Apache /server-status showed CPU as:
CPU Usage: u93.73 s13.02 cu0 cs0 - .0217% CPU load
If PMP means my CPU Load is 23.2 milliseconds then I guess we’re talking mean CPU time for Apache processes, which != load in my understanding or in Apache’s server-status. What am I misunderstanding here?
It would be awesome if we knew what a good/average/bad CPU Load was in terms of the metrics displayed by the Apache monitoring plugin. That’s why I like values in percentages. Milliseconds of CPU time is difficult to interpret when checking the overall CPU load of a service.
