www-apache-bugdb mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@znep.com>
Subject Re: general/1680: Critically anomalous behavior when upgrading from 1.3a1 to 1.3b3
Date Sun, 18 Jan 1998 00:00:05 GMT
The following reply was made to PR general/1680; it has been noted by GNATS.

From: Marc Slemko <marcs@znep.com>
To: Morgan Davis <mdavis@cts.com>
Cc: Apache bugs database <apbugs@apache.org>
Subject: Re: general/1680: Critically anomalous behavior when upgrading   from 1.3a1 to 1.3b3
Date: Sat, 17 Jan 1998 16:47:54 -0700 (MST)

 On Sat, 17 Jan 1998, Morgan Davis wrote:
 > Marc Slemko writes:
 > > Comment out the USE_FLOCK_SERIALIZED_ACCEPT bit from the section
 > > appropriate to your OS in src/main/conf.h. 
 > The server has been running since I last wrote to you with this
 > modification, and have not witnessed the behavior we had seen
 > immediately before with a virgin 1.3b3 (network alarms and pagers
 > going off, stuck children, etc).  By removing that #define, is this a
 > performance hit?  Or just the opposite?  I can't tell if this is a
 > good thing or bad.
 On most systems using serialized accept()s is a performance win.  On
 recent FreeBSD 2.2 releases, you probably get a small performance win by
 not using serialized accepts.  The difference should be quite small under
 the moderate loads you are seeing.
 More importantly, if you use multiple Listen directives then you need some
 sort of accept() locking or things can choke.  If you don't, no worry.
 I'm unsure why you would be seeing this problem with
 USE_FLOCK_SERIALIZED_ACCEPT.  I will have to think about that one.
 > However, the error log file still shows many "mmap_handler" failures,
 Known bug, fixed a few hours after 1.3b3 was released.
 > and more disturbingly periodic rashes of "resource temporarily
 > unavailable - unable to spawn child process" when exec'ing a CGI.
 > Server was handling about 500 hits a minute at the time, with about 45
 > active children.  Basically loafing along with 106 virtual hosts.
 Are you sure you aren't running out of processes or file handles?
 login.conf can trick you very easily.
 What does a "ulimit -a" from the shell where you start Apache show?

View raw message