[squid-users] ICP cache sharing not working after upgrade

From: Len Kranendonk <[email protected]>
Date: Thu, 14 Jul 2005 15:19:28 +0200

After migrating from Squid 2.5.STABLE6 to 2.5.STABLE10 my
two squid servers are not sharing cache anymore.

Both squids are configured in http accelerator mode and one
is a sibling of the other.

In the access_log on 10.0.0.2 I see ICP_QUERIES from the other:

10.0.0.1 - - [14/Jul/2005:15:12:09 +0200] "ICP_QUERY
http://webserver/page.html HTTP/0.0" 0 50 "-" "-" UDP_HIT:NONE

Here the second squid has page.html in cache, but instead of using that,

the first squid has a TCP_MISS and is doing a TIMEOUT_DIRECT.

192.168.1.1 - - [14/Jul/2005:15:15:14 +0200] "GET
http://webserver/page.html HTTP/1.1" 200 104099 "-" "lwp-request/2.06"
TCP_MISS:TIMEOUT_DIRECT

How's that possible ? Has anything changed between STABLE6 and STABLE10
in regard to ICP ?

This is the configuration I'm using for the cache hierarchy:

icp_port 3130
cache_peer 10.0.0.2 sibling 80 3130
acl peers_src src 10.0.0.2 10.0.0.1
acl peers_dst dst 10.0.0.2 10.0.0.1
acl web_access dst webserver
acl web_access_port port 80
cache_peer_access 10.0.0.2 allow web_access
cache_peer_access 10.0.0.2 deny all
http_access allow web_access web_access_port

As for the format of the access_log: I've installed the customlog patch.

Thanks,

Len Kranendonk
Received on Thu Jul 14 2005 - 07:19:32 MDT

This archive was generated by hypermail pre-2.1.9 : Mon Aug 01 2005 - 12:00:02 MDT