Re: ICP parent

From: Duane Wessels <wessels@dont-contact.us>
Date: Tue, 13 Oct 1998 10:47:47 -0600

"Matt Buford" writes:

>I just upgraded to patch2, took the default squid.conf, and added only one
>line:
>
>cache_peer netcache.insync.net parent 80 3130
>
>Then I go to any site:
>
>908280672.194 3673 209.16.31.13 TCP_MISS/200 953 GET
>http://www.digex.net/ - DIRECT/www.digex.net text/html
>908280673.646 1449 209.16.31.13 TCP_MISS/200 9312 GET
>http://data.digex.net/index.cfm - FIRST_PARENT_MISS/netcache.insync.net
>text/html
>908280674.206 1032 209.16.31.13 TCP_MISS/200 2190 GET
>http://data.digex.net/images/sphere.gif - PARENT_HIT/netcache.insync.net
>image/gif
>908280674.554 1363 209.16.31.13 TCP_MISS/200 2780 GET
>http://data.digex.net/images/digex.gif - PARENT_HIT/netcache.insync.net
>image/gif
>908280674.563 1372 209.16.31.13 TCP_MISS/200 655 GET
>http://data.digex.net/images/h-about.gif -
>FIRST_PARENT_MISS/netcache.insync.net image/gif
>
>And as always, the first object retreived after any pause in activity is
>fetched directly instead of through the parent. Any ideas? Shouldn't every
>URL be fetched from the parent except those I have in the stoplist?
>("cgi-bin" and "?" URLs)

This is expected behaviour. There is a "dead_peer" timeout value.
If we don't know the peer has been alive in the last "dead_peer" seconds,
then we assume it is not alive and we don't wait for the ICP reply
to come back.

>Also, after the upgrade and default configuration file, I'm still not seeing
>any server side persistant connections. Is there something special I need
>to do to enable them that is not on by default in the conf file?
>
>Server-side persistent connection counts:
>
> req/
> conn count
> ---- ---------
> 1 140

this doesn't mean its broken. It could mean your workload is just
such that there have been no opportunities to use persistent connections.

do some debugging if you like.

> squid -kdebug
> client -r http://www.netscape.com/
> sleep 3
> client -r http://www.netscape.com/
> squid -kdebug

then look at cache.log for "pconn" and "Keep-Alive" lines.

Duane W.
Received on Tue Oct 13 1998 - 10:45:38 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:42:28 MST