httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <>
Subject Re: somethin amiss on Solaris
Date Thu, 08 Jan 1998 20:25:56 GMT
It it fixed by using USE_FCNTL_SERIALIZED_ACCEPT?  'cause I suspect the
pthreads stuff far more than any of the others... it's the major thing
that's different between solaris and all other platforms.  And it's only
used in 2.5 and later, so wouldn't affect Dirk's 2.4.


On Thu, 8 Jan 1998, Roy T. Fielding wrote:

> >If you hit shift-reload a couple of times and watch the images in the
> >upper right corner of page, you will see that the images on the port 81
> >server takes a long time to finish, while the port 80 server is fine.
> That exactly matches what I am seeing (though I'm getting no response
> from your port 81 server).  I am not using PHP -- just the standard modules
> plus mod_info and mod_status.  In any case, these are plain file requests
> handled using mmap (only 16 syscalls per request).
> I tried testing it with four children and parent all running truss.
> It seems to be starvation of the first child created, since it
> consistently hits the fifth in-line image on a server with 
> StartServers 4
> MinSpareServers 2
> MaxSpareServers 4
> Unfortunately, it is difficult to truss a newly created process
> before the hicup stops.  Hmmm, maybe the problem is in the new
> exponential startup stuff
> BTW, I am also getting only 8 requests/second on a test setup that
> normally handles over 60/second.  This problem is not in 1.2.X
> (I just tested that as well).
> ....Roy

View raw message