Hi Pound people,

I'm using 2.0.1 with the following config on a fairly large/busy site:

LogLevel 4
User "nobody"
Group "nobody"

ListenHTTP
  Address 10.1.230.228
  Port    80

#  ExtendedHTTP 1
#  Server 60

  Service
    Url ".*"

    # t1-1
    BackEnd
      Address 10.1.230.206
      Port    80
    End

    # t1-2
    BackEnd
      Address 10.1.230.180
      Port    80
    End

    Session
      Type  IP
      TTL   300
    End
  End
End

It's been running happily with just one back end node for a while. Since I 
activated the second node in the cluster I've had a few complaints of people 
not being able to log in to the site, which is PHP based with a bit of ajax 
based stuff. Most users have no problem whatsoever. Others log in to the 
front page, then get told they're not logged in like the session data hasn't 
been preserved or pound has decided to swap their node. The problem appears 
to persist for these users, which indicates that they're not being swapped 
round from node to node by pound as I'd expect it to work after the first 
failure.

Is the session stuff in that config ok for persisting on one backend per IP 
for the duration?

Is anyone aware of intermittent problems like this on 2.0.1? Have I missed 
something? I've upped the TTL to 19200 in case that was the issue. As usual 
it's a case of vague reports from external users and nothing reproduceable 
here. Anyone seen similar? Suggestions?

It could of course be an application problem, but with this many variables 
and not being able to find a changelog for 2.0.1 -> 2.0.4  any pointers 
would be greatfully recieved.

--
Tony

On Fri, 2006-04-07 at 15:12 +0000, Tony Howat wrote:[...]

Upgrade to 2.0.4 - 2.0.1 had problems with the session code. Your config
looks OK.[...]

[snip]

Robert Seagall wrote:
[...]

I've gone to 2.0.4 and the problem still occurs. I'm going to add some debug 
code to see if I can see what's going on with respect to session switching.

Now running pound 2.0.4 with only one node configured - though i'm working 
from vague user complaints - but it does seem they stop when in single node 
config. Agh.

Anyone else had problems with PHP sessions with 2.0.4?

--
Tony

Set LogLevel to 3 and see if there are any bad header error messages from
/var/log/messages.  We had a problem with invalid headers being stripped
that caused users' sessions to break in some of our ASP based apps.

Ken

-----Original Message-----
From: "Tony Howat" <arhowat(at)hotmail.com>
To: pound(at)apsis.ch
Date: Sun, 09 Apr 2006 08:54:23 +0000
Subject: Re: [Pound Mailing List] Pound 2.0.1 problem.
[...]


********************************************************************************************
IMPORTANT: The information contained in this message is privileged and
confidential.  It
is intended only for the use of the individual or entity named above.  If the
reader of this message is not the intended recipient, any dissemination or 
reproduction of it is strictly prohibited.  If you have received this
communication
in error, please contact us at postmaster(at)Rezlink.com immediately.  
Thank you.
********************************************************************************************