Re: [squid-users] Parent TCP connection failed

From: Artemis Braja <artemis.braja_at_plus.al>
Date: Mon, 10 Oct 2011 07:31:44 +0000

Thanks Amos for the advices,
It was related to IPv6. After commenting out the ipv6 entry on hosts
file, the parents were successfully connected.

Thanks again!
Artemis

On Sat, 2011-10-08 at 18:19 +1300, Amos Jeffries wrote:
> On 08/10/11 03:51, Artemis Braja wrote:
> > Hello,
> >
> > I'm running two frontends instances on ports 3128 and 3129, and two
> > backend instances on ports 4001 and 4002 on the same box, and
> > loadbalancing between parents using carp.
> > cache_peer localhost parent 4001 0 carp name=backend-1
> > cache_peer localhost parent 4002 0 carp name=backend-2
> >
> > squid -v
> > Squid Cache: Version 3.1.15
> > configure options: '--enable-async-io' '--enable-linux-netfilter'
> > '--enable-storeio=ufs,aufs' '--exec-prefix=/usr' '--bindir=/usr/sbin'
> > '--libexecdir=/usr/lib/squid' '--enable-carp' '--enable-cache-digests'
> > '--enable-removal-policies=lru,heap' --with-squid=/opt/squid-3.1.15
> > --enable-ltdl-convenience
> >
> > After successfully starting all instances I see on the frontends
> > cache.log:
> > 2011/10/07 16:10:56| TCP connection to localhost/4001 failed
> > 2011/10/07 16:10:56| TCP connection to localhost/4002 failed
> > 2011/10/07 16:10:57| TCP connection to localhost/4001 failed
> > 2011/10/07 16:10:57| TCP connection to localhost/4002 failed
> > 2011/10/07 16:10:57| Detected DEAD Parent: backend-2
> > 2011/10/07 16:10:58| TCP connection to localhost/4001 failed
> > 2011/10/07 16:10:58| Detected DEAD Parent: backend-1
> > 2011/10/07 16:11:19| temporary disabling (Service Unavailable) digest
> > from localhost
> > 2011/10/07 16:20:16| TCP connection to localhost/4002 failed
> > 2011/10/07 16:25:20| temporary disabling (Service Unavailable) digest
> > from localhost
> >
> > Because only backends are caching, not connecting to then all traffic
> > from frontends is forwarded to the origin servers, making the storages
> > useless.
>
> Add no-digest to the cache_peer lines on the frontend. That will silence
> the digest messages and make it a little bit longer before DEAD gets
> triggered.
>
> Check that your backend were actually finished loading their caches and
> listening for new traffic before the frontend were connected to them.
>
> Note that Squid-3.1 is IPv6-enabled software so "localhost" is probably
> the IP address ::1. You may need to check your IPv6 firewall settings to
> verify the layers can connect over localhost.
>
> Amos
Received on Mon Oct 10 2011 - 07:31:49 MDT

This archive was generated by hypermail 2.2.0 : Mon Oct 10 2011 - 12:00:03 MDT