httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ben Laurie <...@gonzo.ben.algroup.co.uk>
Subject Re: Apache 1.1.1 after 16 days
Date Tue, 27 Aug 1996 18:22:31 GMT
Alexei Kosut wrote:
> 
> On Tue, 27 Aug 1996, Paul Sutton wrote:
> 
> > Isn't this fixed with the BROKEN_WAIT stuff? It was reported on IRIX and
> > setting #define BROKEN_WAIT gets children reaped every 60 seconds or so.
> > Perhaps this needs setting on Solaris 2 as well?
> 
> Oh, yaeh... that's right. Actually, I think it needs setting all the time.
> I've seen this happen on at least Solaris and BSDI. It probably occurs on
> all systems. I'm more inclined to think it's a bug in Apache, rather than
> in the OS. It's doubtful that (at the least) IRIX, Solaris and BSDI have
> the same, broken, implementation.

I'm inclined to agree, but when I looked into this at the time, I was convinced
that there was no way the "main" wait could reap a child and then fail to
clean it up. The trouble is, I don't see how this is consistent with
BROKEN_WAIT fixing the problem. All it does is wait for each child individually
instead of waiting for the whole lot. Either we have to believe that the
general wait consistently returns the dead child and doesn't clean up, or we
have to believe an OS bug. Unless I missed something.

Don't forget that IRIX, Solaris and BSDI all started with the same code...

> 
> I did spend some time once trying to track down this bug, I never did. I
> think it's related to the timeout code, though. Near as I can tell, a
> request will come in, the process will stop, it will stay alive for some
> amount of time (possibly the length of a timeout) in idle state, then die.
> It's entry is not, however, removed from the scoreboard; therein lies the
> problem, of course.
> 
> I think we should either fix it or permenately enable the BROKEN_WAIT
> code.

I'd still prefer to do this on an OS by OS basis (and preferably check for each
new OS version) because either it is a bug in Apache, in which case we should
fix it, or it is a bug in the OSes, in which case they should fix it. If we
blanket enable it, people won't run into, and it won't get fixed.

Cheers,

Ben.

> 
> -- Alexei Kosut <akosut@organic.com>            The Apache HTTP Server 
>    http://www.nueva.pvt.k12.ca.us/~akosut/      http://www.apache.org/
> 

-- 
Ben Laurie                  Phone: +44 (181) 994 6435
Freelance Consultant and    Fax:   +44 (181) 994 6472
Technical Director          Email: ben@algroup.co.uk
A.L. Digital Ltd,           URL: http://www.algroup.co.uk
London, England.            Apache Group member (http://www.apache.org)

Mime
View raw message