Hi, I have no experience in profiling mod_perl. I'd like to use Devel::NYTProf to profile a production mod_perl backend.
There's an LVS load balancer in front of the mod_perl backends. So far my idea is: - identify 1 backend that I want to execute profiling on - make sure it's idle (weight=0 in LVS) - restart it with -X ?? - assign it some share of the production load - let it run for "n" hours - stop it - collect profiling data Is profiling mod_perl like this at all possible? Does that make sense? So far I've been profiling "statically" by having some production backends write a special log file, letting them run for 1 day, and then analyzing the log file to gather data. This method doesn't take into account stuff like TransHandlers, and doesn't tell me which are the heaviest or most called functions throughout all the code base. So, any ideas? Or rather, what's the state-of-the-art of profiling mod_perl apps? -- Cosimo
