httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "GUMMALAM,MOHAN (HP-Cupertino,ex2)" <mohan_gumma...@hp.com>
Subject RE: tarballs are up
Date Mon, 30 Jul 2001 21:39:13 GMT
-1 for beta

Was able to build and run it on HP-UX on PA-RISC -- with problems.

It still exhibits the same problem as was evident in 2.0.19 and later
releases, there is only 1 process up and running, (apart from the cgid and
the watchdog process).  The configuration I used were as follows:

StartServers         6
MinSpareThreads      5
MaxSpareThreads     50
ThreadsPerChild     25
MaxRequestsPerChild  0

This behaviour on a running system would be as follows.  At peak load, say a
webserver is serving 130 simultaneous request => I have 6 worker processes
running, a total of 150 threads.  At some point later, the load (suddenly)
falls down to say, 90 requests => 60 spare threads.  The current logic would
cause the p_i_s_m() to send a pod, to kill one process.  Since all the
processes are busy serving requests, the process which recd the pod will not
die immediately.  Hence, p_i_s_m() will continue sending pods.  Eventually
all the processes will die (and almost all of them simultaneously).  And
finally, there will be 0 processes running.

Since the MinSpareThreads is 5, a new process will be spawned to meet that
requirement.  And depending on the load, many other new processes will be
spawned to manage it.  So a web-site with variable load, will always see the
process creation and destruction to the extremes - whereas the ideal
behaviour would have been for the webserver to kill exactly 1 process when
the MaxSpareThreads is exceeded.

This is a bug - a bad implementation of the lo-watermark and hi-watermark
logic, and makes Apache web-server 2.0.22 not enterprise quality.  Justin's
patch of two mutexes did fix the problem, but I do not see it being included
even now.

Feel free to correct me if my understanding is fallible, but I'd hate to see
this issue being side-tracked and a release being made nevertheless.

M

> -----Original Message-----
> From: MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1)
> [mailto:madhusudan_mathihalli@hp.com]
> Sent: Monday, July 30, 2001 11:28 AM
> To: 'new-httpd@apache.org'
> Subject: RE: tarballs are up
> 
> 
> Looks good on HPUX/Itanium (IPF)
> 
> -Madhu
> 
> -----Original Message-----
> From: William A. Rowe, Jr. [mailto:wrowe@rowe-clan.net]
> Sent: Monday, July 30, 2001 10:47 AM
> To: new-httpd@apache.org
> Subject: Re: tarballs are up
> 
> 
> Before, and when you announce the beta, try to warn users 
> about the platform
> specific breakages (win 9x/me still don't work, don't know 
> what Unix gotchas
> 
> exist beyond the solaris 7 problem.)
> 
> If we spell out the problems, nobody will be too disappointed.
> 
> ----- Original Message ----- 
> From: "Cliff Woolley" <cliffwoolley@yahoo.com>
> To: <new-httpd@apache.org>
> Sent: Monday, July 30, 2001 12:30 PM
> Subject: Re: tarballs are up
> 
> 
> > On 30 Jul 2001, Jeff Trawick wrote:
> > 
> > > > I just posted the 2.0.22 tarballs at 
> dev.apache.org/dist.  Check 'em
> out.
> > >
> > > +1 for beta
> > 
> > Cooooollll... that's enough +1's for beta (OtherBill, Jeff, 
> myself, and
> > Justin).  That would seem to cover at least Win32 and a few 
> flavors of
> > Unix (at least Linux and Solaris).  I'll wait a little 
> while longer for
> > confirmation from a few other platforms, and then go with it.  =-)
> > 
> > Brian, David, how are OS/2 and BeOS looking?
> > 
> > --Cliff
> > 
> > --------------------------------------------------------------
> >    Cliff Woolley
> >    cliffwoolley@yahoo.com
> >    Charlottesville, VA
> > 
> > 
> > 
> > 
> 

Mime
View raw message