Perrin Harkins <phark...@gmail.com> wrote:
On Sun, Mar 29, 2009 at 4:08 PM, Cosimo Streppone <cos...@streppone.it>
wrote:
Is profiling mod_perl like this at all possible?
Yes.
Does that make sense?
No.
You'll get so much data that you won't be able to make heads or tails
of it. And profiling is heavy enough that people hardly ever do it on
production servers.
Of course. But I could dedicate a spare backend to this for
a reasonable amount of time.
The main problem is that in the past we experienced some kind of
performance problems that only manifested themselves really clearly
in production and only at peak traffic hours.
Out of peak hours, everything was fine.
I'd like to avoid that in the future. In my case, it wasn't
obvious where to look for the root cause of this performance problem.
Just pick one type of request that seems slow, make that request on
your dev machine from your browser, and look at the results. This is
normally all it takes to find bottlenecks.
Yes, this is what we already do.
But it's not enough.
At first, I tried to replicate production load, but it's
nearly impossible. Then I used siege and other tools to replay some
more or less "real" browser sessions, and that's ok.
However, some bad problems start to be evident only with really
high req/s rate.
If you're concerned about finding which things are most frequently
called, just do standard log analysis to find out.
Yes, of course. Thanks.
--
Cosimo
