WEB Server FARMS, clustering and failover:

Let's use the "OSI Layer" here:

Layer 5-7: If you are running Java in a single tiered model, there are cluster based options in the Application server side.

Layer 3.5-4: Network hardware load balancer is an excellent tool. These also allow you do some URL rewriting (the best is a CSS switch; cheap Alteons are available for cheap used). They balance multiple servers for incoming packets and provide failover.

Layer 3: drbd or heartbeat works but not really well. The server, not the Apache process will failover, and it not really scalable.

Layer 3: Ultra Monkey no arp clustering. This is an Alan Cox kernel hack that will allow you to cluster a great number of servers to take a HUGE amounts of hits. If one quits it is removed from the Director.
It works well, and is easy to setup. Ultra Monkey is the name of a product, but the application developed by Alan Cox is a kernel rebuild to route packets based on Mac address rather than IP from and to a Director script/application. Research it and see if you can engineer your own for you distro, or just buy Ultra Monkey.

(503)754-4452 Blackberry || www.obnosis.com
via USB PPP EDVO/CDMA on Dell Latitude PII - Kubuntu 7.10

--- On Thu, 7/3/08, Judd Pickell <pickell@gmail.com> wrote:

From: Judd Pickell <pickell@gmail.com>
Subject: Re: Server failover
To: "Main PLUG discussion list" <plug-discuss@lists.plug.phoenix.az.us>
Cc: klsmith2020@yahoo.com
Date: Thursday, July 3, 2008, 3:40 PM

One of the places I worked for used this: http://bb4.com/ however it
maybe too big for your needs, I am not sure.

On Thu, Jul 3, 2008 at 10:04 AM, Matt Graham <danceswithcrows@usa.net>
wrote:
> From: keith smith <klsmith2020@yahoo.com>
>> I am looking for information on how I might be able to mirror one
>> server onto another and set them up so if the main server dies
>> the second server will take over

 automatically.
>
> heartbeat and drbd.  http://linux-ha.org/ and http://drbd.org/ to
> start.
>
>> The mirroring would require the website content itself and the
>> MySql data.
>
> drbd works below the filesystem level, if you want to use that.  You
> really need a dedicated high-speed link between the 2 systems if you
> want to use drbd.  Unless you use one of the newer and less well
> tested filesystems such as GFS, if you use drbd, you have to have the
> primary machine mount the filesystem, and the secondary machine can't
> mount the filesystem at all.  Scripts allow the secondary machine to
> become primary if/when the primary flails over.
>
>> Now if I had 3 servers would I be able to do this by having them
>> all three working as a "farm" to serve up multiple websites?
>
> Not in the way described above.  drbd is very
 much
 a 2-machine system.
> You could have a drbd cluster with large disks acting as the main
> fileserver, and 1..N machines that mount the filesystems on that
> server via NFS.  Then machines 1..N could serve content via apache
> or apache combined with ldirectord.
>
>> Is it possible to do this on a small budget?
>
> Define "small".  You need boxes, time to set it all up, and
disks.
>
> --
> Matt G / Dances With Crows
> The Crow202 Blog:  http://crow202.org/wordpress/
> There is no Darkness in Eternity/But only Light too dim for us to see
>
>
> ---------------------------------------------------
> PLUG-discuss mailing list - PLUG-discuss@lists.plug.phoenix.az.us
> To subscribe, unsubscribe, or to change your mail settings:
>

 http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss
>
---------------------------------------------------
PLUG-discuss mailing list - PLUG-discuss@lists.plug.phoenix.az.us
To subscribe, unsubscribe, or to change your mail settings:
http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss