Re: [squid-users] fail-safe and load balancing with reverse proxy

Dear Amos and Dean,

Thank you for your answers.

I think, these timeout settings only reduces the failed attempts but doesn't
solve the problem.

Is there a correct solution for full fail-safe peer selection?

In my opinion I do this with ICP queries.
Can I do perfect fail-safe peer selection with two proxies between ICP?

reverse-proxy <--> proxy with 2 ips <--> web server

Best regards,
László Király

---------- Original Message -----------
From: Amos Jeffries <squid3_at_treenet.co.nz>
To: squid-users_at_squid-cache.org
Sent: Thu, 03 Jun 2010 02:10:33 +1200
Subject: Re: [squid-users] fail-safe and load balancing with reverse proxy

> Dean Weimer wrote:
> > Try using "peer_connect_timeout" You can lower the time out so it fails
over faster.
> >
> > Thanks,
> > Dean Weimer
> > Network Administrator
> > Orscheln Management Co
> >
> >> -----Original Message-----
> >> From: Király László [mailto:kiru_at_mail.madalbal.hu]
> >> Sent: Wednesday, June 02, 2010 3:14 AM
> >> To: squid-users_at_squid-cache.org
> >> Subject: Re: [squid-users] fail-safe and load balancing with reverse proxy
> >>
> >> Okay, I compile a brand new 3.1.4 squid with --icmp-enable option.
> >>
> >> I added also to the squid.conf:
> >> ---
> >> pinger_program /usr/local/squid/libexec/pinger
> >> query_icmp on
> >> test_reachability on
> >> ---
> >>
> >> It didn't help. :S
> >>
> >>> Hi List,
> >>>
> >>> I use a squid3-3.0.STABLE8 reverse proxy on a debian system.
> >>> It makes forward queries to web server, which is accessible from 2
> >>> public ips.
> >>>
> >>> My peer config:
> >>> ---
> >>> cache_peer x.y.z.57 parent 80 0 no-query no-digest no-netdb-exchange
> >>> originserver name=parent1 round-robin login=PASS weight=16
> >>> cache_peer a.b.c.118 parent 80 0 no-query no-digest no-netdb-exchange
> >>> originserver name=parent2 round-robin login=PASS weight=1
> >>> ---
> >>>
> >>> I would like to do a fail-safe connection to the web server.
> >>>
> >>> It's working, but if one of the public ips isn't accessible, there
> >>> is some Connection timed out (110) proxy message until the parent is
> >>> detected as dead, while the proxy tries to query the offline parent.
> >>>
> >>> How can I eliminate this thing?
> >>> Why squid doesn't resend the query to the another parent?
> >>>
> >>> I cannot set ICP queries while the parent is a simple web server.
> >>> Is there a way to make better dead peer detection?
> >>>
> >>> Can I do this whith icmp queries?
> >>>
> >>> Best regards,
> >>> László Király
> >> ------- End of Original Message -------
> >
>
> 3.1 also brings in the cache_peer options connect-timeout and
> connect-fail-limit to set per-peer how long to wait for TCP
> connections to failover and how many attempts to allow before
> declaring it dead.
>
> Note: ICMP and NetDB is used to detect the closest source when
> multiple are available. Not the live/dead status of a peer.
>
> Amos
> --
> Please be using
> Current Stable Squid 2.7.STABLE9 or 3.1.4
------- End of Original Message -------
Received on Thu Jun 03 2010 - 08:32:35 MDT