RE: [squid-users] "Impossible keep-alive header"

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Sat, 24 Jan 2004 02:30:09 +0100 (CET)

On Fri, 23 Jan 2004, M J wrote:

> Sorry Henrik, I use one old topic that had the same problem / configuration
> as I do, but I now a realize that there is a lot of them, I m sorry. I ll
> give details. I was having a problem that the file cache.log was full of
> these warnings
>
> 2004/01/23 19:38:38| ctx: exit level 0
> 2004/01/23 19:38:38| ctx: enter level 0:
> 'http://applications.atpco.net/extranet/coresystems/pdf/fm_rules_n3_b.pdf'
> 2004/01/23 19:38:38| httpProcessReplyHeader: Impossible keep-alive header
> from
> 'http://applications.atpco.net/extranet/coresystems/pdf/fm_rules_n3_b.pdf'

As this is a restricted server I can not verify if the warning is valid or
not.

> and I used one patch that you sent (keep-alive.patch ) and the warning
> decreased from 1 warning per 10 seconds to 1 warning per 4 or 5 hours
> (acceptble).
> Today I have compiled squid-2.5.STABLE4-20040123.tar.bz2 to know if the
> problem was solved but the cache.log file starts to increase those warnings.

2.5.STABLE4-20040123 should be identical to your previous sources. There
has been no changes after the keep-alive.patch (2004-01-19).

Cronological history of recent Squid-2.5 changes:

2004-01-19 18:46 hno

* src/HttpMsg.c: Spelling correction

2004-01-19 18:34 hno

* src/HttpMsg.c: Bug #890: Various HTTP workarounds and minor
  corrections

  Squid is still HTTP/1.0 so we can not assume a HTTP/1.1 message
  indicates persistent connections.

2004-01-15 08:43 hno

* src/acl.c: Corrected debug level in detection of empty
  proxy_auth acls

2004-01-15 08:40 hno

* src/acl.c: Refuse to accept empty proxy_auth acl declarations. In
  combination with the splay_splay unpredictability on empty splay
  trees this could give very strange results.

2004-01-15 08:38 hno

* lib/splay.c: Make splay_splay predictable if called on empty
  splay trees.

2004-01-14 19:47 hno

* src/: http.c, structs.h: Bug #890: Some servers send broken
  keep-alive headers (IIS-5)

  The initial patch was not a little to broad and claimed fully
  valid body-less replies violated keep-alive. (304, HEAD etc)

2004-01-12 11:51 hno

* src/: client_side.c, http.c, structs.h: Bug #890: Various HTTP
  workarounds and minor corrections

  - Automatically time out incorrectly signalled persistent
    connections after 10 seconds of inactitivy. Also gives a warning
    in cache.log

  [...]

Regards
Henrik
Received on Fri Jan 23 2004 - 18:30:14 MST

This archive was generated by hypermail pre-2.1.9 : Sun Feb 01 2004 - 12:00:08 MST