httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject Re: Our recent Apache woes (fwd)
Date Fri, 27 Jun 1997 22:18:19 GMT
I am 100% convinced that serialized accepts are the way to go.  I'm still
convincing myself that weakly serialized (i.e. a maximum of N tasks inside
the loop) is a good thing as well (consider SMP boxes). 

The problem described below regarding the load is probably still explained
by the lack of serialization.  The difference is probably that hpux uses a
different formula for calculating its load avg than sunos 4.1.x.  I'm not
sure how to relate the rest of the problem though. 

We could start defining one of the serialization directives in 1.2.1
according to the reports we've received for:  osf, hpux, sunos.  Use
whatever the users have reported works for them.  (Or whoever uses those
architectures could tell me which of fcntl or flock works for you.) 

Dean

On Fri, 27 Jun 1997, Jason A. Dour wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> 
> Here's a letter from the UofL webmaster regarding our recent problems
> upgrading our old Apache server for www.louisville.edu to version 1.2.0.
> 
> For background information, since it's not provided in his note, our
> machine is an HP, running:
> 		HP-UX homer B.10.10 E 9000/819
> 
> He compiled it with HP's compiler, and received no warning or errors.
> Aside from a standard config, he's using suEXEC and a DCE/kerberos5 module
> that was developed here at UofL.
> 
> Can anyone shed a little more light on the situation for us?  We have it
> working, but the starving sockets fix wasn't the clearest answer...
> 
> Any information/help would be greatly appreciated, guys.
> 
> Thanks,
> Jason
> # Jason A. Dour <jad@bcc.louisville.edu>                            1101
> # Programmer Analyst II; Department of Radiation Oncology; Univ. of Lou.
> # Finger for URLs, PGP public key, geek code, PJ Harvey info, et cetera.
> 
> - ---------- Forwarded message ----------
> Date: Fri, 27 Jun 1997 10:59:33 -0500 (EDT)
> From: Ken Hagan <khhaga01@homer.louisville.edu>
> To: "Jason A. Dour" <jadour01@homer.louisville.edu>,
>     "Donald S. Teiser" <dsteis01@homer.louisville.edu>,
>     Keith Stevenson <ktstev01@homer.louisville.edu>,
>     "Michael H. Dyre" <mhdyre01@homer.louisville.edu>,
>     webmaster@homer.louisville.edu
> Subject: Our recent Apache woes
> 
> Hello,
>   We have recently been undergoing the seemingly simple migration 
> from Apache 1.0.5 to Apache 1.2.0.  During the migration, we 
> have had some outtages of the web server.  They are hopefully 
> resolved, so I thought I would send a note detailing what I think
> was wrong.  
> We fired up Apache 1.2.0 on Friday, 13 June.  I now realize how silly
> it was to do such a thing on Friday the 13.  It ran fine until the 
> next Thursday sometime very early in the morning. There was no responding 
> web server at 0730 when people started coming in to work.  The 
> processes that were running we just sitting there laughing at you when 
> you connected and made a request (hopefully the chortles will be less 
> irritating in Apache 2.x).  We killed all of the processes and restarted 
> the server.  It died later that day and proceeded to die several more times
> during the next week or so.
> We tried many different things to no avail.  Removing extraneous modules, 
> removing our DCE/Kerberos5 auth module, recompiling many different ways.
> None of these worked.  We were suffering from the load spike problem
> documented in the Apache Bug Database, but not in the same way as was 
> documented (the bug report listed instantaneous spikes and our seemed to 
> go up quickly, but trickle down).  The bug report on load spikes had a 
> pointer to another bug report concerning starving sockets.  That 
> bug report is 
> http://www.apache.org/bugdb.cgi/full/467
> The bug report doesn't make any specific references to our plight, but 
> the fix described therein seems to have fixed our starving sockets and our
> load spikes.  The odd thing is that the bug report was for servers using 
> multiple listen directives and we are not. But anyhow, the change was made
> to the makefile, the server recompiled, and now the site has been up for 
> almost a day (at a 100,000+hits I might add).  I just thought I would send 
> a report of our woes to keep everyone briefed.
> 
> 
>  Ken.
> *****************************************************************
> Ken Hagan                          	Programmer Analyst 
> khhaga01@homer.louisville.edu      	Research and Development
> (502)852-1701                      	Information Technology
> http://www.louisville.edu/~khhaga01/    University of Louisville
> *****************************************************************
> 
> -----BEGIN PGP SIGNATURE-----
> Version: 2.6.2
> 
> iQCVAwUBM7Pdo5o1JaC71RLxAQEcowP8DJanCFrUKWXD15Yk6/LD5/AUQogHpInC
> ctDH51Nb4ysRimxDOtXc4o9oceKsGDdM2JWWfhCrJvYiVVN3tZROTgMNvBFScBtJ
> sSC/fRgJ9oyxhunFbd9yrFzY7W4s1XDnb/IHqB6y04oo35pcONUbwR41KfYBYrqs
> rotQE9iM9Lw=
> =aVxl
> -----END PGP SIGNATURE-----
> 
> 


Mime
View raw message