[wellylug] CPU load

[Pete Black]

The load average is the number of processes waiting to be serviced by 
the kernel. Often, high load is caused by swapping, or applications 
leaking memory and causing swapping.

Anything that spawns a lot of processes that are all looking to get 
concurrent access to a resource will tend to raise the load average, and 
a high load average does not necessarily imply a lack of responsiveness 
- e.g. i have an app that spawns 500 processes all of which are 
requesting data from the network - this tends to blow my load average 
out to 20 or above, but there is no lack of responsiveness with other 
apps on the machine.

As the machine is definitely chugging, I would guess that swap is the 
issue, and that one or more of your processes is using vast amounts of 
RAM. Run top and press 'M' (Shift-m) to sort by memory usage to see if 
you can see anything. Also processes sitting in 'D' uninterruptible 
sleep for long periods are indicative of problems.

-Pete


>Quoting Gordon Paynter <lists at paynter.info>:
>  
>
>>run top, I can see one process that registers in the %CPU column: it
>>gets 
>>99.9%, even though it is niced down to 10 (i.e. it is low-priority).
>>    
>>
>How about telling us what that process is?
>As for the load average: it's made up of both
>CPU and I/O ... so if you had that box being a
>bridge between two high-speed interfaces that
>may be one thing to investigate, or if the hdd
>was running without dma and MySQL was quite busy ...
>
>
>Cheers,
>Andrej
>
>
>  
>