httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@znep.com>
Subject Re: lingering close
Date Sun, 19 Jan 1997 04:25:32 GMT
On Sat, 18 Jan 1997, Ben Laurie wrote:

> Ben Laurie wrote:
> > 
> > Randy Terbush wrote:
> > 
> > I certainly agree that a second shutdown is at best unneeded, so I think we
> > should try it to see if it fixes anyones problems. I'm afraid I've not been
> > paying too close attention to the state of play on this - do we have any sites
> > where we can test stuff?
> 
> It also occurs to me that doing a lingering close by default is wrong, and
> may well be the cause of the performance drop that has been observed.

It shouldn't have anything to do with performance of _a_ client reading,
but if you are hitting it very hard with multiple connections then it will
take longer for the children to get back into the pool to service another
request which could slow things down a little I guess.

> 
> We should find which OSes need it and only do it on those.

I think I agree.  From what I read, it was only put in for certain broken
clients and should not be needed in general.  Some SVR4 (and SCO 5?)
systems are all I noticed in the archives that were mentioned as needing
it...

Not having to use lingering_close would solve many of the FIN_WAIT_2
problems, but not all.  Sigh.


Mime
View raw message