[SQU] [Fwd: Advice for CARP in SQUID]

From: Michael Ngo <[email protected]>
Date: Mon, 25 Sep 2000 13:47:18 +1100

--
------------------------------------------------------------
Michael NGO,
Bachelor of Arts/ Bachelor of Engineering,
The University Of Melbourne,
Victoria 3010, AUSTRALIA
phone: +61 3 8344 9195
Email: mngo@ee.mu.oz.au
------------------------------------------------------------

attached mail follows:


Sorry for not making myself clearn, but yes, my query was about CARP in
SQUID.
  Thanks for the advice with the directive "prefer_direct", now that I have
turned this off the client
squid can communicate with the parent. However, as I have noticed in the
access.log of the client squid,
it reads FIRST_UP_PARENT instead of CARP, meaning that the client retrieved
the object form the first
parent in a list of parents and did not apply the CARP hashing algorithm to
select the parents.

How do I configure the client squid so that requests are load balanced
amongt the two parents???

Kind regards,

  Michael.

Henrik Nordstrom wrote:

> Is it Squid we are talking here, or something else?
>
> If it is Squid then Squid only implements the hasing based request
> routing part of the CARP client, with all routing weights hardcoded in
> squid.conf on each and every client cache. The parents does not actually
> have to run CARP at all with the current Squid configuration.
>
> If you are using CARP without ICP then you most likely must also use
> prefer_direct off.
>
> The Squid implementation is only about 1/4 of a full CARP
> implementation. Things missing are:
> a) Fetching+Parsing of a CARP membership table, to not have everything
> hardcoded in squid.conf.
> b) Intra-Array routing of requests (optional, low prio)
> c) CARP master which builds and publishes the CARP membership table for
> clients to use.
>
> --
> Henrik Nordstrom
> Squid hacker
>
> Michael Ngo wrote:
> >
> > Hi Henrik,
> >
> > I've just mailed a question regardin CARP to the web caching
> > mailing list, but I've noticed that you
> > have answered a similar question before. Like I outlined in the
> > group mail, my problem is one that
> > after running the configure script to enable carp and configuring one
> > of three boxes (emu1) to use
> > two other boxes (emu2 and emu3) as parents, I've found that they are
> > not talking to each other. That is,
> > all requests on emu1 are forwarded straight to the original server and
> > are not retrieved through its
> > parent caches.
> >
> >
> > /--emu2 (configured with CARP) as parent
> > emu1 -----X
> > \--emu3 (configured with CARP) as parent
> >
> >
> > I hope you can give me some advice on this.
> >
> > Thanks in advance.
> >
> > Michael.
> >
> >
> > --
> > ------------------------------------------------------------
> > Michael NGO,
> > Bachelor of Arts/ Bachelor of Engineering,
> > The University Of Melbourne,
> > Victoria 3010, AUSTRALIA
> > phone: +61 3 8344 9195
> > Email: mngo@ee.mu.oz.au
> > ------------------------------------------------------------
> >
> >
>
> --
> To unsubscribe, see http://www.squid-cache.org/mailing-lists.html

--
------------------------------------------------------------
Michael NGO,
Bachelor of Arts/ Bachelor of Engineering,
The University Of Melbourne,
Victoria 3010, AUSTRALIA
phone: +61 3 8344 9195
Email: mngo@ee.mu.oz.au
------------------------------------------------------------
--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
Received on Sun Sep 24 2000 - 20:49:01 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:55:27 MST