(Resend with required Subject field)

I'm seeing fairly frequent diagnostic messages of the sort:

   error copy server cont: Connection reset by peer

   error flush headers to <IP ADDR>: Connection reset by peer
 
   error read from <IP ADDR>: Input/output error

   error copy client cont: Connection timed out

   error copy client cont: Success

   error copy server cont: Broken pipe

It would be helpful to know what might be causing these errors.  Some are
to be expected, I know, but the frequency is high enough to make me
suspicious of some other problem.

A bit of background:

We are using Pound as a reverse proxy without SSL connections.  Version is
Pound-current (aka 1.8 candidate) of Aug 13, OpenSSL 0.9.66, RH7.3. I have
659 UrlGroups in my config file.

Occasionally we have seen an explosive expansion of the number of pound
instances in the process map which has occasionally required that we
restart pound.

I am seeing some significant performance problems somewhere in the
Net/Pound/Zope/MySQL path even on a mostly unloaded machine.  Pound does
not appear to be sopping up significant CPU time.

On Thursday 26 August 2004 20.05, Dennis Allison wrote:[...]

As part of the great memory leak hunt we have introduced much more detailed 
logging. These messages just indicate at which point during the response the 
client browser closed the connection.

As a side note: the latest -current is from Aug. 23rd, so you may want to 
update. I promise after 1.8 we'll go to sub-minor (1.8.1, 1.8.2, etc) numbers 
to make tracking easier.[...]