Re: Strange Message

mneto@buriti.com.br writes:

>Hi Squiders!
> I am running a Linux 2.0.27 with squid 1.0.22 and I found these messages
>in my /var/adm/syslog. What is this Failure Ratio at 1.01 and
>hit-only-mode ? Is this correct ?
>
>Thanks, Mario.
>
>Jan 20 09:38:35 baco squid[83]: [20/Jan/1997:09:38:35 -0400] icp.c:250:^I
>Failur e Ratio at 1.01
>Jan 20 09:38:35 baco squid[83]: [20/Jan/1997:09:38:35 -0400] icp.c:251:^I
>Going into hit-only-mode for 5 minutes...

/*
 * This function is designed to serve a fairly specific purpose.
 * Occasionally our vBNS-connected caches can talk to each other, but not
 * the rest of the world. Here we try to detect frequent failures which
 * make the cache unusable (e.g. DNS lookup and connect() failures). If
 * the failure:success ratio goes above 1.0 then we go into "hit only"
 * mode where we only return UDP_HIT or UDP_RELOADING. (UDP_RELOADING
 * isn't quite the appropriate term but it gets the job done). Neighbors
 * will only fetch HITs from us if they are using the ICP protocol. We
 * stay in this mode for 5 minutes.
 *
 * Duane W., Sept 16, 1996
 */

The failure ratio is the ratio of failed requests to successful
requests. A failed request is one that is logged with
ERR_DNS_FAIL, ERR_CONNECT_FAIL, or ERR_READ_ERROR.

When your cache starts returning as many failures as successes, it
sort of disables itself for 5 minutes. This could happen if there
is a problem with your upstream provider, but all your child caches
can still reach you.

Duane W.
Received on Tue Jan 21 1997 - 08:19:39 MST