Re: linux problem ["James H. Cloos Jr." ] (fwd)

[Brian Behlendorf <br...@organic.com>]

Can we *please* have someone using Linux, who can recreate this problem, 
find out exactly where in the steps from 0.8.9 to 0.8.11 this behavior 
started showing up in Linux?  I.e., recreate each level by the patches 
submitted and implemented.... I'm not sure this occurs with 0.8.10 so 
that's why I thought I would bring it up.  Failing that, is there a way 
to bring this to a Linux networking guru's attention?  Having a 
repeatable failure condition would be very useful.

I won't be able to get to it this weekend (I'll be in the middle of 
black rock desert in nevada with about 5000 other people burning a 6 
story high wooden man and parasailing behind a car and other fun 
activities :), but next week I'm going to try turning that list of 
requirements for a test suite I gave the other day into a system of perl 
scripts and standard test config files.  

	Brian

On Wed, 30 Aug 1995, Brian Behlendorf wrote:
> ---------- Forwarded message ----------
> Date: Thu, 31 Aug 1995 01:11:51 -0500
> From: "James H. Cloos Jr." <cl...@io.com>
> To: new-httpd@hyperreal.com
> Subject: Re:  linux problem
> 
> Robert> Hmmm... this *could* be some kind of scheduler anomaly, with
> Robert> long-lived server processes --- has anyone tried setting
> Robert> MaxRequestsPerChild to a low value, and seeing if that changes
> Robert> things?
> 
> I tried that this afternoon.  I cut the MaxRequestsPerChild from 30 to
> 6, and got about 1600 sockt failures and about 100K hits in the next
> 12 hours.  This compares with 2388 socket errors and 170K hits in 24
> hours before the change.
> 
> Doesn't seem to have made much of a change.
> 
> -JimC
> -- 
> James H. Cloos, Jr.	Snail:	POBox 18122 Austin, TX 78760-8122   Usenix,SAGE
> cloos@io.com		URL:	http://www.jhcloos.com/~cloos/      LPF
> 

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com  brian@hyperreal.com  http://www.[hyperreal,organic].com/