You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Maurizio Marini <ma...@datalogica.com> on 2002/10/29 17:35:46 UTC

Semaphore Arrays

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello

Redhat 7.3, apache 1.3.23-14 and the rest by distribution

we are noting in a production env. that  after stopped, apache deny to start 
again for shortness of Semaphore Arrays
the default by distro is 128 arrays for a total of 32k semaphores

the strange thing is that apache consumes 1 sem for each array
so, after 128 sem, the count is exhausted

we are wondering , then, if this is a known bug solved in next apache 
releases;
we are installing 1.3.27, we hope in this one the problem is fixed

we would add that this is a rev-proxy web server, so mod_proxy and mod_rewrite 
are intensively used
10x in advance for your help
- -- 
Maurizio Marini
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE9vrji4Q/49nIJTlwRAkJTAJ4iUAHfMI2+5muFLEivFReudS5OIACfRgf7
xqzvCLfPympxZVOf96sYQMQ=
=uZZR
-----END PGP SIGNATURE-----

Re: Semaphore Arrays

Posted by Jeff Trawick <tr...@attglobal.net>.
Jeff Trawick <tr...@attglobal.net> writes:

> You can switch to another mutex mechanism ("AcceptMutex fcntl" in
> httpd.conf) to work around this problem if your Apache is consistently
> crashing or you're using a 3rd party module which allocates sems and
> doesn't clean up.

uhh, my last comment was pretty stupid...  AcceptMutex isn't going to
make a 3rd party module do anything different 

-- 
Jeff Trawick | trawick@attglobal.net
Born in Roswell... married an alien...

Re: Semaphore Arrays

Posted by Jeff Trawick <tr...@attglobal.net>.
There are no known bugs in this area (no reason to expect improvement
by upgrading Apache).  Note that if Apache is not stopping cleanly the
semaphore won't be cleaned up.

You'll need to manually clean up the semaphores.  Use "ipcs -s" to
display ids on Linux and "ipcrm sem <id>" to remove sem on Linux.

You can switch to another mutex mechanism ("AcceptMutex fcntl" in
httpd.conf) to work around this problem if your Apache is consistently
crashing or you're using a 3rd party module which allocates sems and
doesn't clean up.

Maurizio Marini <ma...@datalogica.com> writes:

> Hello
> 
> Redhat 7.3, apache 1.3.23-14 and the rest by distribution
> 
> we are noting in a production env. that  after stopped, apache deny to start 
> again for shortness of Semaphore Arrays
> the default by distro is 128 arrays for a total of 32k semaphores
> 
> the strange thing is that apache consumes 1 sem for each array
> so, after 128 sem, the count is exhausted
> 
> we are wondering , then, if this is a known bug solved in next apache 
> releases;
> we are installing 1.3.27, we hope in this one the problem is fixed
> 
> we would add that this is a rev-proxy web server, so mod_proxy and mod_rewrite 
> are intensively used
> 10x in advance for your help
> -- 
> Maurizio Marini
> 

-- 
Jeff Trawick | trawick@attglobal.net
Born in Roswell... married an alien...