Noob question about server load

jhadley

All of the Linux utilities I know of work with 1, 5 and 15 minute load averages, but I am working on something that could benefit from having a shorter load average, or even reading the load number rather than an average. Is this possible?

Alex_LiquidHost

You are looking for something to monitor the load in real time?

jhadley

Yes

Alex_LiquidHost

Doesent top monitor in real time?

jhadley

@Alex_LiquidHost said: Doesent top monitor in real time?

No, top gives averages over 1, 5 and 15 minutes.

Alex_LiquidHost

Might be, I am not sure. However isn't this the load at the current time: http://i.imgur.com/XmqId.png

And the load average at the top for 1, 5, 15 mins?

MrAndroid

uptime calls getloadavg (avg, 3);

However, that function can only maximum return 3 (being 1,5 and 15). You could probably make your own formula to get the CPU usage, IO usage and memory usage and output a result yourself however.

jhadley

@Alex_LiquidHost said: However isn't this the load at the current time: http://i.imgur.com/XmqId.png

No, that's only CPU usage, which is just one component of the load.

NickW

The only potentially useful thing I found. http://stackoverflow.com/questions/8879370/reading-a-shorter-than-1-minute-load-average-on-a-linux-system

The other suggestion was hacking at the kernel.

jhadley

No easier way at all?

Mon5t3r

what do you expect from average?

gsrdgrdghd

@Mon5t3r said: what do you expect from average?

uptime returns the average over 1, 5 and 15 minutes, i guess @jhadley also wants the average over e.g. 10 seconds or so.

budingyun

What other component of the load? :o

Mon5t3r

@gsrdgrdghd said: uptime returns the average over 1, 5 and 15 minutes, i guess jhadley also wants the average over e.g. 10 seconds or so.

sorry that comment should be in number 3, below jhadley's "yes" comment.. i forgot hitting post comment button hahaha

@jhadley well maybe /proc/uptime can help you out. also take a look on this site

NickM

The load average is a rolling average of the the number of processes in the run queue divided by the number of processing cores. The first column of vmstat's output is the run queue length. So you could use vmstat|tail -1|cut -d" " -f2 to get a point-in-time snapshot of the run queue, then divide it by the number of cores. If you're consistently getting a number higher than the number of cores you have, that would probably indicate a problem.

NickW

@NickM said: The load average is a rolling average of the the number of processes in the run queue divided by the number of processing cores.

It is not divided by the number of processing cores/threads, which is why this number should be taken into account whenever looking at a load average.

NickM

@NickW is correct. Sorry for the confusion.

ErawanArifNugroho

how about vpsinfo?