[squid-users] Problem with squid / Plumtree Portal

Hello all
 
I'm having a very strange issue with our squid cluster
and our Plumtree Portal.
 
I'm running Squid version 2.4 stable 6. We have a
three node cluster hardware load balanced which has
been in production for 2 + years. It is running on
Compaq DL 380 G2s with 1 GB RAM, 32 GB Raid 5 HDs and
Intel 10/100 NICs at 100 FD. An average load at peak
week day is approx 10,000 users. Due to some broken
internal web sites (some = many) we are running DNS
processes locally rather than pointing DNS to our
internal DNS servers directly, which brings us to the
problem...
 
A couple of months ago we started seeing forwarding
loop errors in the cache.log indicating issue with our
Plumtree Portal environment - extract as follows:
 
2005/03/04 16:23:03| WARNING: Forwarding loop detected
for:
GET /ptcs/gadgets/admin/gadget.jsp HTTP/1.0
Host: X.X.X.X:9080
 
(Note that the true IP of the host has been changed to
X.X.X.X)
 
This error was being generated several hundred times a
day, coupled with DNS failure messages, seemingly due
to the DNS processes filling up behind the squid
process hanging on this error.
 
I've worked with the development team on the portal to
set all assocated servers to bypass the proxy for all
internal transactions, but I'm still getting the error
on their external transactions.
 
I've been all through the docs / FAQ etc for squid,
but the only reference I see for forwarding loops
points to a mis-configuration issue with the visible
versus unique host names, which is set correctly on
all nodes. I also assume that this would show up for
all transactions, not just those associated with our
portal environment.
 
Thanks in advance for any ideas on this, and please
let me know what other detail may help to point to the
issue
 
Fred Clark

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
Received on Tue Mar 22 2005 - 11:37:59 MST