httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Ames <grega...@raleigh.ibm.com>
Subject Re: Can I take down apache.org?
Date Wed, 14 Feb 2001 22:01:22 GMT
rbb@covalent.net wrote:
> 
> With Bill's most recent changes, I believe we are ready to try to put 2.0
> on Apache.org.  Would anybody mind if I took down apache 1.3 on Apache.org
> and put up apache 2.0 before 8:00 tonight?  I know this is probably a bad
> idea, but I am excited about the idea that this might work.
> 
> Somebody just say, NO!, and I won't do it.
> 
> Greg, what are the chances you would be willing to build on apache.org
> immediately?

We still have a bug where sometimes a server will loop infinitely in "W"
state, according to server-status.  I think that's what you & I were
seeing Monday nite.  I was able to reproduce it late yesterday afternoon
on :8092 with less destructive effects - two processes were looping in
"W" state, and I was able to kill them easily but didn't get doc :-( 
You should have an offlist email I sent about the problem.

Paul Reder tried to reproduce it on his system; no luck, but he is
seeing the lingering close bug I think.  (Paul: "cvs up" will make you
happy, I predict)

Here's a plan:  I will "cvs up" apache.org, build, and start it on
:8092.  Once that's done, let's see if it's a good time to try the log
replay test that triggered the loop against 8092, and do it & get doc if
we hit the bug.  I could use some help with this piece so we don't
impact users more than necessary.  If we don't hit the loop bug on 8092
again, then I'm OK with putting it into production, but want to stand by
to watch.  

If we do hit a loop, we want truss and a core dump of a looper.  

btw, let's create a new state in the scoreboard for lingering close, and
give it a unique letter in server-status.  "W" is overloaded.

Greg

Mime
View raw message