On Sun, Sep 05, 2004 at 08:31:57AM +0200, Nick Wilson wrote:
>
> * and then Jon LaBadie declared....
> > > to allocate to AMANDA. Here is the output of df -h
> > >
> > > Filesystem Size Used Avail Use% Mounted on
> > > /dev/hdb1 187G 18G 169G 10% /
> > > /dev/hda3 37G 5.8G 31G 16% /opt
> > > none 506M 0 506M 0% /dev/shm
> > > /dev/sda1 187G 33M 187G 1% /mnt/extDrive
> > >
> > >
> > > Ideally, i'd like some of the ext drive set aside for other stuff..
> > >
> >
> > how much data do you have to backup?
>
> The entire contents of /dev/hda3 and /dev/hdb1 (see above) I just want a
> complete restorable backup done of the entire system.
Not a complete answer; data and disk capacity are not the same thing.
I can see hdb1 is 187GB and you currently are using 10% of it for data.
What do you anticipate for tomorrow and next year?
BTW it is quite unusual to have your entire system (minus /opt)
on a single humongous root file system. More often it will be
subdivided into logical boundries, eg. /tmp with short-term, not
to be retained across boot, /var with logs and admin stuff,
/home with user data, /usr/local with stuff you add, /export
with shared data, ...
Imagine a year from now looking for that one file you need in
a 150GB backup of / as currently configured.
>
> > how frequently will you backup?
>
> Nightly, at 3am
>
> > how many full dumps vs. incrementals will you want to keep?
>
> Probably overwrite the backups each night. (incremental?)
>
Bzzzzt, bad answer :(
Imagine a powerfailure happening at 3:20am. It corrupts your
humongous root drive but your external drive seems ok.
Unfortunately, you have now destroyed yesterday's backup as today's
was underway and you don't have today's yet as it is not complete.
The common way to handle backups is periodically dump everything
and then for a number of days backup only what has changed. This
is refered to as a "dumpcycle", one full backups and a set of
incrementals. It is also common to retain more than one dumpcycle,
often several.
The size of your data, not disks factors into how much space to
leave for amanda's full dumps. Also, how rapidly your data
changes affects how big your incrementals are.
> > will you compress or no?
>
> Sure, why not...
>
It factors into how much space your backups will take.
Different data compresses by different amounts. Some might
actually expand slightly or at least not compress significantly.
Others may shrink to just 10% of the original size.
> > where will you have a holding disk if any?
>
> Not sure what you mean? - I have an ext. drive that I want to use for
> this purpose (200gb)
Amanda was designed to handle many "client" systems on a lan backing
up to a single "server" system with a tape drive. Only one stream
can be going from the server to the tape drive. But amanda can
handle many streams from clients by using a "holding disk" as a
buffer until the data can be taped.
You are going to use a hard disk instead of a tape drive. But
amanda will treat it just as if it were a tape drive. We call
them "file tapes" or "virtual tapes (vtapes)".
>
> > is this the only system you will backup?
>
> Yes.
Because you are backing up just one host, the server,
the holding disk is not so important.
>
> > mixing virtual tape, holding disk, and data being backed up
> > on the same HD spindle is a sure performance hit.
>
> Yeah, but im planning on having it do its stuff while I sleep, i dont
> run any public services from this machine, it's just my work/play box
> ;-)
>
> Thanks very much for the help, i can see my original question really
> wasnt very good so thanks for helping me get the relevant facts out!
So it becomes a formula of how big is one full dump, how many extra
incrementals in a dumpcycle, how big is an incremental, how many
dumpcycles do you plan to retain, and how much will your data compress.
You currently have about 25GB of data. If it might grow to 40GB,
use that number. Guess on the change of data rate, ?10%?. How
long between full dumps, ?1 week? equals six incrementals equals
6x40x0.1 or 24GB. Thus a dumpcycle would be about 64GB of backups.
How many dumpcycles to retain ?1.5?, ?2?, ?3?. If the latter,
then you will have on disk 3x64GB (192GB) of backup data. How
much will it compress ?20%?, ?50%?, ?70%?. Fifty percent is
a commonly used guess, so you would need about 100GB.
Of course, that is using the factors I picked, you pick your own.
jl
--
Jon H. LaBadie jon AT jgcomp DOT com
JG Computing
4455 Province Line Road (609) 252-0159
Princeton, NJ 08540-4322 (609) 683-7220 (fax)
|
