[consulting] Distributed Loads, Failover and Dynamic Content

[Leigh]

Its not super tricky but you gotta know your way around a couple of things

main reason we love it that we can make a new virtual server and host 
new setups on there
without requiring new hardware (other than sometimes a little more RAM)

AND we can move entire _running_ virtual machines between different 
servers if something
needs upgrading or something dies, and the failover machine picks up 
from exactly the
same state it was sent from the main machine

(A) Get the hang of Xen (we use 3.0) and get drupal running on a 
virtual machine

(B) Setup a spare/failover Xen machine

(C) If the server needs upgrading, use the xen commands to "send" the 
virtual machine to
another hardware (your spare/failover Xen machine) and send back again 
when the original
server is up

(D) underpants gnomes

(E) profit


Should start a drupal book/chapter on hosting methods

Quoting John Sechrest <sechrest at jas.peak.org>:

>
>
> I would really like to hear the details of how
> you have the Xen system set up.
>
>
>
> leighm at linuxbandwagon.com writes:
>
> % I run a bunch of apache/php servers under the linux/xen virtual system
> % platform.
> %
> % it works really, really well, i can move running images of a machine to
> % another computer
> % if there is a failure.
> %
> % also you could probably have two machines running the same xen instance
> % for extra
> % reliability somehow?
> %
> % Quoting Kieran Lal <kieran at civicspacelabs.org>:
> %
> % >
> % > On Jul 14, 2006, at 5:41 PM, Hugh Esco wrote:
> % >
> % >> I'm working with folks who would like to physically distribute
> % >> their server load so that they might be prepared for peak demand as
> % >> well as to provide for fail-over as part of a disaster recovery plan.
> % >
> % > You'll have to deploy different strategies for peak demand versus
> % > disaster recovery.
> % >
> % > For disaster recovery you can dedicate a server to slave
> % > replication.  MySQL Lead Architect, Brian Aker,  and MySQL senior
> % > performance engineer Peter Zaitev have been discussing in public.
> % >
> % > Brian: http://krow.livejournal.com/434562.html
> % > Peter: http://www.mysqlperformanceblog.com/2006/07/07/thoughts-on-
> % > mysql-replication/
> % >
> % > For peak load you'll want to do what Allie suggests which is multi-
> % > master replication.
> % >
> % > This is a good overview of MySQL replication.  But be aware, drupal
> % > is not read versus write aware yet so typical slave configuration
> % > won't work.
> % >
> % > http://jcole.us/mysqluc/2006/MySQL%20UC%20-%20Replication%20for%
> % > 20Scaling%20and%20High%20Availability.pdf
> % >
> % > There is a good article that details how to set auto increments
> % > between master slaves offset by 10 so that you can achieve MySQL
> % > master replication.  I can't find it right now but will do so and
> % > post shortly.
> % >
> % > Do you plan to have a lot of files as well?
> % >
> % > Cheers,
> % > Kieran
> %
> %
> %
> %
> %
> % --
> % Roses are #FF0000,
> % Violets are #0000FF
> % all of my base,
> % are belong to you
> %
> % ----------------------------------------------------------------
> % This message was sent using IMP, the Internet Messaging Program.
> %
> % _______________________________________________
> % consulting mailing list
> % consulting at drupal.org
> % http://lists.drupal.org/mailman/listinfo/consulting
> %
>
> - - -----
> John Sechrest          .         Helping people use
>                        .           computers and the Internet
>                          .            more effectively
>                             .
>                                 .       Internet: sechrest at peak.org
>                                      .
>                                              . http://www.peak.org/~sechrest
>
> - ------- End of Unsent Draft
>
> ------- End of Forwarded Message
>
>



----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.