Re: [squid-users] what debug-section should i use.

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Thu, 5 Aug 2004 00:01:16 +0200 (CEST)

> 2004/08/04 18:05:51| comm_accept: FD 9: (53) Software caused connection
> abort
> 2004/08/04 18:05:51| httpAccept: FD 9: accept failure: (53) Software caused
> connection
> abort

These are harmless and can be ignored unless you can connect them with
specific customer complaints.

> - i have read about the users hitting the stop button before the tcp
> handshake but i am pretty sure that this is not my reason. Is there any
> other explanation for this entries ?

None that I know of. Well, there is ofcourse many other situations when
the same thing happens at the network layer. The presumed error is that
the client aborted the request before even the TCP connection was accepted
by Squid.

A tcpdump of the traffic at the time may be useful in confirming this is
not your reason. If this is the reason then you will close to when this is
logged in cache.log see the following sequence:

   1. SYN from client to Squid port
   2. SYN+ACK from Squid port
   3. ACK from client to Squid port
   4. Maybe a few data packets from client to Squid port
   5. Maybe a few ACK packets from Squid port to client
   6. A TCP RESET (RST) from client to Squid port

If you instead see

   [as above up to and including 5]
   6. FIN from Squid port to client

then the problem is something else.

Make sure to have your Squid running in debug mode to allow you to relate
cache.log messages with IP addresses and ports in the tcpdump data, if not
it is pretty much a wild guess to try to relate the two.. (the failed
connection should be seen in the tcpdump between two successful ones
logged in cache.log)

Regards
Henrik
Received on Wed Aug 04 2004 - 16:01:21 MDT

This archive was generated by hypermail pre-2.1.9 : Wed Sep 01 2004 - 12:00:01 MDT