Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Re: Backup & Recovery Procedures
What management thinks of when they think about the database is different
from what we think of. Management wants the data to be available to users.
So you may want your backup plan to consider where replacement hardware will
come from if necessary, what to do if the database is lost because of
network failure, telephone failure, microwave failure, electrical failure.
Then you have to consider where you will get the people to fix everything
and how to reach them. Are you up 24/7? Can a DBA always reach the site
within 2 minutes?, 2 hours? You want to be sure your support contract with
Oracle has not lapsed.
As for how often backups are tested, Sybrand is right. The most common
answer is never. I've worked at big companies that refused to test their
backups because they didn't want to make a server even temporarily available
for testing! I was called in to one site where it was found that the person
responsible for taking the cold backup every night had simply stopped doing
it four months earlier.
Van
"Horvath Andras" <raas_at_balu.sch.bme.hu> wrote in message
news:8r1meg$hut$1_at_goliat.eik.bme.hu...
> Backup. Well.
>
> Then, create a copy of your databases :-) and simulate disaster.
>
> What can you lose?
> - data, possibly by user misbehaviour (!)
> - datafiles in non-system tablespace
> - datafiles in the SYSTEM tablespace w/ data dictionary and things
> - control file
> - log file (you DO write logs to at least 2 different disks, dont't
> you?)
> - archived log file(s), more or less recent
> - init.ora's, scripts used to do this and that including the backup
> - whole database ("rm -rf /")
> - whole disk (hw faliure) (you DO have at least 2 control files..)
> - whole machine (fire, flood, thieves, whatever).
> - install media (Oracle CD, op.system CD and such)
> - more or less recent backups themselves (!!), lost or corrupted
> - the BCP itself :->
>
>
>
> Think this is all of 'em. PUT DOWN what to do in which case. Try and go
> through all so that when it comes to real disaster time and the whole
> management is banging your head you don't have to think much but do
> things as written.
>
> UPDATE the BCP regularly as your system changes.
>
> Another two questions are:
> - should you be able to restore a full-working 1-year-old backup of the
> DB or yesterday's backup will be enough?
> - how much TIME is allowed to be spent on recovery (eg in case of a
> disk faliure) ? Ten minutes, ten hours or ten days?
> Use a watch when testing your BCP.
>
> I'm preparing this plan right now for my new employer:) and would be
> glad if the more experienced out there in this newsgroup commented the
> above.
>
> Regards
> raas
> --
> "With enough free time, effort, and an input string that looks like the
> winning entry in last year's obfuscated-C contest, you can turn
> someone's simple mistake into a nationally syndicated news story."
Received on Fri Sep 29 2000 - 13:53:05 CDT
![]() |
![]() |