[Web4lib] Library Web - Backup Plan and Server Setup

[Francis Kayiwa]

On Feb 7, 2007, at 12:34 PM, Jonathan Blackburn wrote:


>
> This (primarily) includes 2 LAMP machines:
>
> 1. One hosting our main website, which is powered by the open- 
> source CMS
> Drupal
>
> 2. An "auxilary" webserver (still in its infancy), which hosts  
> important but
> slightly-less "mission critical" functions like our Libraries'  
> blogging
> platform (WPMU).

Curious what's WPMU

>
> Any possible hardware limitations have yet to be decided, but we  
> have at
> least 2 new machines (yet to be installed) to work with right  
> now . . .
>
> Our desired set-up would support:
>
> 1. Frequent backups (daily - best practices here welcomed?)

dump(8) -provides a fantastic strategy on planning for backup

http://tinyurl.com/2gnguh

YMMV.

>
> 2. Failover with "minimal intervention" - We are specifically  
> worried about
> nights and weekends when there may not be staff immediately  
> available to
> troubleshoot problems.

Not sure I follow. If hardware goes down at night/weekend... I  
wouldn't trust not being there to bring them back up. I will concede  
that I am missing something. :-)

http://backuppc.sourceforge.net/

Might meet your needs however. It can backup over the web... never  
used it.

>
> 3. Convenient method for copying over changes from development  
> environment
> to production - These include software upgrades and settings  
> changes that
> may require restart and/or possible downtime.

cvs; subversion?

>
> As you may have guessed, load balancing is much less of an issue,  
> though I
> would certainly not ignore any performance improvement tips you  
> chose to
> offer.
>
> Any thoughts?  I am very interested in how others of you have set  
> up your
> servers to promote optimum redundancy and active development.
>
> We have some ideas ourselves, but I am intentionally "casting the  
> net wide"
> to see what others of you out there have done.

I have always used the following maxim for backup strategies.

How much down time can I afford? Test this.
Nothing like finding out that your best strategy is rubbish (as I  
recently found out sadly). So this rule is huge. In this particular  
case the machine itself could afford a significant amount of downtime.

Given the above. Take whatever you are backing up and try to rebuild  
it on a different machine and see how fast you would bring it back  
up. If that time is reasonable and meets you expectations then you  
got a good system going and I would stick with that.

Then decide what disaster you are preparing for. Are we worried about

-user error(s)
-sysadmin error(s)
-0wned box
-Dead disk (shudder)
-dead (insert other part of machine)
-Fire/ Water acts or supreme being
-tactical nuclear strike (hey -you never know :-))

depending on which of these you are worried about try to decide how  
you will be back online. In your situation you mention they are  
webservers. If this was a debian machine (you mentioned LAMP) I would  
place  /var and /etc as the most important partitions. I would place  
less emphases on all other partitions assuming you have the  
aforementioned CVS/subversion etc.,

HTH

./fxk
===============
Francis Kayiwa
Library Systems Team
4-180, MC 234
T: +1.312.996.2716
W: http://www.uic.edu/~kayiwa
Key: http://tigger.uic.edu/~kayiwa/kayiwa.gpg