Thanks for your excellent resource.
As a stop-gap (for now) the following script is cron'd every minute on the
server to log basic load profile and to email `top` snapshot if things get
out of hand.
-----monitor-----------
uptime>>/jobs/monitor.log;
temp=$(cat /proc/loadavg);
if [ $(echo -e "scale=0 \n ${temp%% *}/0.01 \nquit \n" | bc) -gt 100 ]
then
file="/jobs/monitor.peak_`date '+%m-%d-%y-%H:%M:%S'`";
`top -b -n 1 > ${file}`;
mail root -s "Load Warning: ${temp}" < ${file};
fi
----------------
Better ideas are always appreciated.
Bill
On Thursday 12 May 2005 16:57, Stas Bekman wrote:
> Bill Whillers wrote:
> > Hi,
> >
> > I'd really appreciate any ideas or "best practices" you might have before
> > re-inventing with Yet Another Benchmarker/Profiler...
> >
> > We are expecting an unusually high amount of traffic over the next month
> > on one of our Apache/MySQL machines and I'd like to profile this in a
> > simple, "bootstap" way (i.e. process logger running every 2 minutes)
> > capturing only critical info on the server (CPU, IO, Apache, MySQL
> > processes, etc.).
> >
> > I'm hoping that this log process will be very light-weight.
>
> Bill, please try: http://modperlbook.org/html/part2.html
