[squid-users] tcp errors

Hi all,

I have squid running on a Debian Sarge Linux server, acting as a caching
proxy. When squid is initially started nothing much happens (as it
should), but once squid has been accessed by a client machine I will get
HEAPS (about 6MB/hour) of tcp traffic which is just errors and retries.
Here is a sample of the errors taken from a tcpdump. Port 8000 is
another squid proxy server at my ISP, my clients are accessing my squid
proxy at myServer:8080 ....

[TCP Dup Ack 598#1] 8000 > 1759
8000 > 1760 [Ack]
1760 > 8000 [Ack]
[TCP Previous segment lost] 8000 > 1760
[TCP Dup Ack 613#1] 1760 > 8000 [ACK]
[TCP Retransmission] 8000 > 1760 [ACK]

I realise that this doesn't sound like a squid problem but the server
can directly access the upstream proxy without errors (eg: doing a wget
directly to port 8000 on the upstream proxy rather than localhost:8080).
And my local clients can use a squid proxy running on another server
without problems. The only consistency in the error tcp data is my local
squid being run and used at least once.
Oh, when running, squid will fetch the wanted webpage although slowly.

Any ideas on what this problem may be or how to debug it?

thanks
David...
Received on Thu Jan 06 2005 - 20:41:40 MST