[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: backup plan bare metal

hi ya rodney

On Tue, 20 Sep 2005, Rodney Richison wrote:

> Would be interested in seeing what some of you use for a backup plan. 

backup plan ... i always run 3-5 duplicate servers at the same time ..
	just change the ip# and the backup machine would be live

> Tar?

find | tar | grep -iv is all what one needs to do 8-day, 32-day, 180-day
incrementals + full weekly backups

> Easy/quick way to restore bare metal?

take a virgin disk .. plug it into any pc and have a go at it

once the initial systemis installed from cdrom or you-favorit-tool
	- it can be 1 minute or 1hr or 1 day to get a working box
	  back to the same state as your master box

	- apply all patches ( my rules, even if it makes it newer than
	the original master server )

- restore your /home data onto the newly recreated machine from bare metal

> Rsync?

rsync is extremely bad idea because:
	- whatever caused your main server to die will propagate
	itself on your backup server when you use rsync

	- rsync does NOT save a copy of the file before you overwrite it

> Can it do bare metal?

anything can do bare metal .. once you have a workign system such
that you can run "rsync"

> what about hard links?

hard links, soft linke and hanging links and anything else is
all part of the choice of different apps you use to create probalems
or avoid problems because  foo-app you used cannot do the trick

> Again, easy/quick way to restoe entire debian server?

- boot a minimal deb install from fopppy or cdrom 
- do net installs and upgrades
- restore your /home data and other directories *-you-* modified

> Unfortunatly, I've found mondo unreliable for bare metal. Though when it 
> works, it's the cat's meow..

80% - 90% of most of the backup programs will all fail one or more fo the
backup and restore tests ... for any number of reasons ..

	- you figure out what is important to you

	- why did the primary server die
	- do you care if the problem propagates to other servers
	- do you care that the new replacement hardware is different
	- do you want to fiddle with the backups daily, weekly, monthly ??
	- do you need a gui to point-n-click around

	- do you need to recover the data of the machine till Jul 4, 2005
	  before the first sign of the cracker and sleeper that you just
	  found a month or 3mon or a year later

	- is automated backup that works by itself with 1 line in cron
	  or point-n-clicky backups

	- how fast does your backup need to go live when the primary goes

	- is automatic fail over a good or bad idea ??
		- the carcker will rm -rf your backup too if you're nto
		careful about services and passwd and vulnerabilities

	- how much USER data of backup do you need
		10MB, 100MB, 1TB, 10TB ...

	- how do you protect against dead disks, dead fans, dead
	powersupplies, dead ethernet card, ...

	- how do you protect against kernel upgrades and other
	required system patches and upgrades

	- how much is it gonna cost ( in hourly $$$ losses ) if the
	  primary server dies

	- how much is the backups gonna costs and where is it 

	- gazillion metrics

== all those answers will dictate which backup and (baremetal) restore you
== should be using

c ya

Reply to: