httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject Re: [PATCH] Brian's nph stuck in graceful bug
Date Sat, 04 Oct 1997 02:31:00 GMT


On Fri, 3 Oct 1997, Brian Behlendorf wrote:

> At 06:21 PM 10/3/97 -0700, Dean Gaudet wrote:
> >The child that invokes the nph cgi does time out properly, but then gets
> >stuck waiting on the child forever because of a waitpid() in the
> >cgi_handler.  This problem exists in all previous versions of apache as
> >well.  But they have no way of solving it -- since they have no idea how
> >much traffic has gone over the nph socket they don't know if it should be
> >timed out.  The nph cgi really has to deal with the timeout. 
> 
> Right; I guess the only reason I'm seeing it now is that the nph stays
> bound to the child in our current model, rather than simply giving the
> socket to the nph script and then moving on to the next request.  Curious
> though that I didn't see a process table full of nph- scripts earlier...

You didn't start using USR1 until recently, and prior to that you used
HUP, right?  With HUP the httpd child is killed off aggressively, it isn't
with USR1.  Even in the pre-1.3 code the child always hangs around, it
never goes off to the next request ... 'cause it needs to write the log.

Dean


Mime
View raw message