high "Load Average"

[Stefan Parvu]

On 12:34 Fri 12 Mar     , Sessna wrote:
> I am using nginx + php-fpm (quad-core PC) + MySQL (on another PC) and facing an interesting issue:
> after some time nginx is running, system "load average" goes up (sometimes up to 30 and even more points) while CPU usage is still less then 30% With such a high load system becomes very slow and unresponsive. Restarting nginx helps, but only for a little while. Any ideas on the matter?
>

You should check CPU%, Disk IO% and probable look for 
operations on the wait queue. sysrec does this on Solaris, 
Im working to port the recorder on Linux, FreeBSD. 
You should somehow count for yourself:

#           CPU            # usr + sys time across all CPUs
#           Memory         # free RAM. freemem from availrmem
#           Disk           # %busy. r+w times across all Disks
#           Network        # throughput. r+w bytes across all NICs
#
# Saturation,
#           CPU            # threads on the run queue
#           Memory         # scan rate of the page scanner
#           Disk           # operations on the wait queue
#           Network        # errors due to buffer saturation


Load average is a metric which might not really tell you a very useful image 
unless you really know what is means and how to interpret that. Check it out:

 http://perfdynamics.blogspot.com/2007/04/how-long-should-queue-be.html
 
I would start with:

 - Disk IO, CPU% and look for signs of saturation.
 - cpuplayer (is your machine spending more time in SYS/USER ?)
 - useful to collect longer trends in order to see whats going on


Hope it helps,
Stefan

---
References:

 sysrec:
  http://systemdatarecorder.org/recording/recorders.html

 cpuplayer:
  http://systemdatarecorder.org/cpuplayer/