Phoenix.pm: stack backtraces from lockedup programs

Scott Walters scott at illogics.org
Fri Dec 5 14:55:33 CST 2003


User mode thread packages and some kernel mode thread pages implement
threads as processes. Even though memory is shared between the processes,
it shows up once in the process table for each thread, so Apache is
losing its own threads there. Report it, wait for the next version,
but like Linux 2.4, Apache 2 is going to have some serious growing pains
before it matures.

(Perhaps someone will bite onto this flamebait since they seem to be 
ignoring the rest...)

Perhaps you can run FastCGI or Apache 1.3 on another port, and
set Apache 2 to proxy requests for URLs which map to CGIs or
your CGI? 

-scott

On  0, doug at phoenixinternet.com wrote:
> 
> > Oooooh, Apache 2. Apache 2 is threaded. You wouldn't beleive the volume
> > of traffic that the problems related to this are generating...
> >
> > When this happens, er, when this message appears in the log, is
> > process 4239 still in the process table? Can you see it? Is
> > it zombie status, or ready to run, or running (is it a multiple CPU
> > system?), or what? In other words, is Apache on crack, or is the
> > process actually hanging? Are the headers correct when you run it
> > from the command line?
> >
> > -scott
> 
> Apache actually ends up killing the processes, but I have noticed after
> testing this repeatedly, that there are an abnormaly high number of Apache
> processes running.  So, I think that the processes are locking up, Apache
> spawns new ones, and then gives me warnings about the locked up ones when
> it kills them before restart.  That's my theory anyway.
> 
> <SNIP>



More information about the Phoenix-pm mailing list