httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Randy Terbush <ra...@zyzzyva.com>
Subject Re: Apache 1.1.1 after 16 days
Date Wed, 28 Aug 1996 13:41:08 GMT
> On Tue, 27 Aug 1996, Alexei Kosut wrote:
> > On Tue, 27 Aug 1996, Michael Smith wrote:
> > > I have had apache 1.1.1 running on our solaris 2.5 machine for 16 days
> > > without it going down.  Below is the output from /status.  As you can
> > > see (by looking at the SS column) there are quite a few connections
> > > which are still up but look like they've been just sitting around for a
> > > few days (I have TimeOut 1200 set in httpd.conf).  What is the meaning
> > > of this - is it my fault (oh please tell me it's not!), or is it
> > > something in the server?
> >
> > It's something in the server. We've seen it around occasionally, and no
> > one really knows why. Somehow, the parent is losing touch with its
> > children. They die, but without clearing their slots in the scoreboard.
> > Hopefully, someone will fix it one of these days.
> 
> 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?
> 
> Paul
> UK Web
> 

Actually, I'm not seeing a problem on Solaris with unreaped children.
I am however seeing a problem where the scoreboard seems to not update
the time properly. This could be a sign of the same thing. 

I glanced at the code in question a few days ago and raised the question
about a race when writing the the scroreboard info. Jim indicated
that he had a hunch. Did that pan out Jim?

I am seeing the problem consistantly on Solaris, but not on FreeBSD.





Mime
View raw message