Re: [squid-users] Local port number logging woes

2015-01-22 Thread Carl-Daniel Hailfinger

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

Hi Amos,

thanks a lot for your answer.

On 22.01.2015 05:22, Amos Jeffries wrote:
 On 22/01/2015 7:00 a.m., Carl-Daniel Hailfinger wrote:
  I'm using cascaded/hierarchical Squid instances, one per machine.
  [...]

  This works great except for one oddity: Quite a few Squid log
  entries of the child proxy have 0 as local port number of the
  last server or peer connection. I have absolutely no idea why that
  would be the case. AFAICS cached entries have - as local port
  number and that's fine because it means there was no associated
  parent proxy connection.

  Am I doing something wrong? Are there any cases where the log
  format code %lp would legitimately yield 0?

 It is possible for connections which were never completely setup. The
 server/peer details will all be known, but the local TCP details will
 not be set by completion of the TCP handshake.

Ah, that may explain it. So this would only happen if the TCP connection
to the parent proxy can't be established completely.
A colleague of mine thought it might be related to parent proxy
connection keepalive. I'll forward your answer.

Thanks again!

Regards,
Carl-Daniel
-BEGIN PGP SIGNATURE-
Version: GnuPG v1

iD8DBQFUwS1oRdNMz2eF/AERAv8kAJsF0XuTwBJvcy74MRpFrdJ7P5klQwCgnrg6
LEP4EkBarBc9qByymqo1hD0=
=fgvz
-END PGP SIGNATURE-


___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


[squid-users] Local port number logging woes

2015-01-21 Thread Carl-Daniel Hailfinger
Hi,

I'm using cascaded/hierarchical Squid instances, one per machine. To get
the ability to correlate access.log entries between instances, I have
extended the default squid log format by %p and %lp .
For correlation, I use the following two sets of information:
The parent proxy uses URL, timestamp, client source port.
The child proxy uses URL, timestamp, local port number of the last
server or peer connection.
Even with some slight timing variations due to caching/lookups/network,
the child proxy local port number of the last peer connection and the
parent proxy client source port help tremendously matching those entries
against each other.

This works great except for one oddity: Quite a few Squid log entries of
the child proxy have 0 as local port number of the last server or peer
connection. I have absolutely no idea why that would be the case. AFAICS
cached entries have - as local port number and that's fine because it
means there was no associated parent proxy connection.

Am I doing something wrong? Are there any cases where the log format
code %lp would legitimately yield 0?

Regards,
Carl-Daniel
___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


Re: [squid-users] Local port number logging woes

2015-01-21 Thread Amos Jeffries
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

On 22/01/2015 7:00 a.m., Carl-Daniel Hailfinger wrote:
 Hi,
 
 I'm using cascaded/hierarchical Squid instances, one per machine.
 To get the ability to correlate access.log entries between
 instances, I have extended the default squid log format by %p and
 %lp . For correlation, I use the following two sets of
 information: The parent proxy uses URL, timestamp, client
 source port. The child proxy uses URL, timestamp, local port
 number of the last server or peer connection. Even with some
 slight timing variations due to caching/lookups/network, the child
 proxy local port number of the last peer connection and the parent
 proxy client source port help tremendously matching those entries 
 against each other.
 
 This works great except for one oddity: Quite a few Squid log
 entries of the child proxy have 0 as local port number of the
 last server or peer connection. I have absolutely no idea why that
 would be the case. AFAICS cached entries have - as local port
 number and that's fine because it means there was no associated
 parent proxy connection.
 
 Am I doing something wrong? Are there any cases where the log
 format code %lp would legitimately yield 0?

It is possible for connections which were never completely setup. The
server/peer details will all be known, but the local TCP details will
not be set by completion of the TCP handshake.

Amos

-BEGIN PGP SIGNATURE-
Version: GnuPG v2.0.22 (MingW32)

iQEcBAEBAgAGBQJUwHr6AAoJELJo5wb/XPRj1e0H/R732wuWifxxJjiQuGQJxaej
vc8c07NS02slvpVSDvLNjyOAMPl++oBgCEvy4EBHiWkR1MP/N7z4Utj4FS1c1HQA
rEJqA0UCe8cnm6V/8zjrEd3l+N+zYx5oj/P4Yv97TmQlmLeMLeVZyaeJcOfMarKo
yS9Tdyy7NIuQioTzFKfZ/1mclJeAKJ1fY2QLjzIWzekDh0NNvmX/3W3NEuFf1fsR
cuYQ4vbTliCbkuugtvmJm/GOATAl3jNvP+pJp+5s1JPe+dYFlqj2All/eR2stai4
ElJaHQm5f2GgU3g/kuZvSHiXuMkMEmwvGj+pwlSu5+Okdi0XpMKx03BnRqPPJTQ=
=SKrI
-END PGP SIGNATURE-
___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users