[Samba] Is Samba Shadowcopying can be used in Production Environement with more than 20 TB of data

Chuck Kollars ckollars9 at yahoo.com
Thu Feb 7 20:25:49 GMT 2008

> ... there will be more than 20TB of data to be 
> backup weekly which will take lots of hours. ...

Check out the `rsync` spinoff of Samba. 

`Rsync`s basic idea is "copy what's changed" rather
than just copying everything. It does so very well and
very quickly. Copying only changed files can easily be
a couple of orders of magnitude quicker than copying
the whole thing.

The possible flaw with this strategy that used to keep
people from implementing it was that the determination
of what's changed had to be _perfect_. A backup's no
good if it only contains 99% of the current data. The
`rsync` tool provides the needed reliabilty, making
this strategy possible in real life rather than just
pie in the sky. 

(Of course your backup medium needs to be a disk farm
rather than tapes...)

My situation is much smaller than yours: a little over
1000 users with a total of a little over 100GB of
data. When I started using `rsync`, my backups went
from many hours once a month (clearly not frequent
enough, but we couldn't afford to do better) to ~10
minutes every day. (I don't use any features of Samba
itself, and I don't use any aspect of LVM.)

And that ~10 minutes is even with the backup on a
separate machine accessed over a network, so
bandwidth's limited to 100MB. A SAN would probably do
quite a bit better. (The completely separate machine
is our way of avoiding a single point of failure.)

(And because the backup is to another disk, the backup
disk can be made available read-only to lots of folks.
As a result, in my situation anybody can restore any
individual file at any time virtually

(The first time will of course take a long long time,
but after that daily updates will be real quick.)

-Chuck Kollars

Looking for last minute shopping deals?  
Find them fast with Yahoo! Search.  http://tools.search.yahoo.com/newsearch/category.php?category=shopping

More information about the samba mailing list