Re: Squid 1.2b19 on NLANR caches broken

From: Duane Wessels <>
Date: Sun, 29 Mar 1998 23:28:34 -0700

"Jordan Mendelson" writes:

> a popular auction house is always causing uc.cache or
>pb.cache to give a 'Connection refused' message, however when I force my 1.1
>server to get it directly, these don't happen.

I didn't happen "always" but it did happen occasionally. It appears to
be a strange interaction with persistent connections. Squid is receiving
ECONNRESET's on some persistent connections. e.g.:

21:25:46.580408 surf.16444 > . ack 12674 win 61440
21:27:33.401131 surf.16444 > P 150:299(149) ack 12674 win 61440
21:27:33.465228 > surf.16444: R 24326162:24326162(0) win 0

Anyway, Squid has now been fixed to handle this case more gracefully and
the new version is running on our caches. Please let us know if you
notice further problems.

> is the same way.
>The odd part is that I would have thought the retry patch would retry any
>errored fetches from the parent caches, but it doesnt.
>Microsoft's Active Setup for Outlook 98 caused all sorts of problems with
>the Squid 1.2 servers. A parent server said a file it was downloading was
>1.8 megs when in reality it was 2.1 megs. So, I opened it up to a site which
>wasn't cached on one of the NLANR caches and tried getting it again. This
>time i hit cancel to force an abort. I tried getting the file again from the
>same place, low and behold the file was truncated on the NLANR cache.

doesn't happen for me

surf-wessels ~ 440> grep TogetherAgain.mp3 $CL/store.log
891239086.802 RELEASE 01000118 200 891239058 887082894 -1 text/plain 4808916/89600 GET
891239102.894 RELEASE 0100011C 200 891239072 887082894 -1 text/plain 4808916/192512 GET

surf-wessels ~ 441> grep TogetherAgain.mp3 $CL/access.log
891239086.802 6285 TCP_MISS/200 89866 GET - DIRECT/ text/plain
891239102.894 8011 TCP_MISS/200 192266 GET - DIRECT/ text/plain

>GameSpy, a popular internet game server browser has a bug which makes it not
>put the 'HTTP/1.0' after the 'GET /URL' request. Squid doesn't understand
>this and forces an error. Unfortunatly, a lot of my users use this program.
>Is there a way to disable this behavorior?

Not yet.

Duane W.
Received on Sun Mar 29 1998 - 22:31:33 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:39:29 MST