Amanda-Users

Re: What tapes can I send off-site ? How do I implement this backup strategy ?

2005-05-10 16:43:33
Subject: Re: What tapes can I send off-site ? How do I implement this backup strategy ?
From: Matt Hyclak <hyclak AT math.ohiou DOT edu>
To: amanda-users AT amanda DOT org
Date: Tue, 10 May 2005 16:24:14 -0400
On Tue, May 10, 2005 at 03:45:57PM -0400, Guy Dallaire enlightened us:
> We use a nice GUI program to do that, but the product is outdated and
> buying a new licence is too costly. Amanda seems good, and we use more
> and more open source software here so we decided to try it out.
> 
> Our current method of backup is the traditional incrementals Mon, Tue,
> Wed, Thu, and Friday we do a Full backup. No backups on week end.
> 
> The tapes are stored in a locked cabinet outside the server room.
> After 3 "business days" on site, the tapes are sent to a remote
> location.  This permits to have the tape handy if we have to restore a
> file that was modified and deleted in the last 2 days.
> 
> Monthly tapes (Full backup) are kept forever for archiving/financial
> regulation  pupose.
> 
> I'm having a hard time figuring out how to do something similar with amanda.
> 

You'll need to keep "dumpcycle" worth of tapes on site. If you set dumpcycle
to 1 week, and runspercycle to 5 (M-F), then you'll need to keep the latest
5 tapes to be able to do those quick restores. What a lot of people will do
with enough tapes is have 4 "sets" of tapes. In your case you might have
something like the following:

* Tape1, Tape2, Tape3, Tape4, Tape5 - "current" set that backups are performed
  on

* Tape6, Tape7, Tape8, Tape9, Tape10 - "most recent" set from last week,
  stored in your locked cabinet

* Tape11, Tape12, Tape13, Tape14, Tape15 - set stored offsite

* Tape16, Tape17, Tape18, Tape19, Tape20 - in transit back onsite to become
  next weeks "current"

Depending on how accessible your offsite location is, you can probably get
away with 3 "sets"

> I'm not familiar with backup levels. It's a concept I always had a
> difficulty to grasp.  If I send the tapes older than 3 days to the
> remote site, will the tapes of the last three days ALWAYS contain the
> file modified in the last 3 days ?
>

Backup levels work like this:

Level 0 contains every file
Level 1 contains every file that has changed since the last Level 0
Level 2 contains every file that has changed since the last Level 1
Level n contains every file that has changed since the last Level n-1

For a simple example, let's consider some directory "foo". "foo" contains the
files "a", "b", "c", and "d".

* Monday - Level 0 - contains a,b,c and d
- User edits file b
* Tuesday - Level 1 - contains b
- User edits file c
* Wednesday - Level 1 - contains b and c
- User edits file a
* Thursday - Level 2 - contains a

So the idea is that if you need to restore an entire directory, you use the
latest-dated of each level, in this case Monday, Wednesday and Thursday's
tapes.

Amanda uses the bump* variables to decide when to increase to the next
level. Above it seems that the level was bumped arbitrarily from 1 to 2, but
what happened for example was that b and c were both extremely large files,
and a was very small. It wouldn't make much sense to use tape for those
large files when the only change is the small "a" file, so it bumps the
level up.

> How can I manage to do full backups of every server each FRIDAY ? I
> know I should create a new config specifying always full and an
> infinite tapecylce and run it on fridays, and run the "regular"
> schedule on the other days.
>

It is generally not recommended to do so. Why do you feel it necessary to
only do full dumps on Friday? Let amanda even out the schedule and you will
find good usage accross tapes, and fairly consistent backup runtimes.

> One last question: Is there an easy way with amanda to do an "ad hoc"
> backup ? I mean, say I have to backup a group of directories on a
> server, do I have to create a new config, a new disklist, etc... ?
> There must be an easier way ?
> 

The host would have to be in the disklist, but you can do individual
backups. amdump takes optional host and disk parameters, so if a machine was
offline during the backup, the next morning you can bring it up and run
something like:

amdump CONFIGNAME missinghost /backupdir

> I'm reading the documentation, but there is a lot of doc, and not
> enough samples and use-cases IMHO.
> 

I hope my examples helped.

Matt

-- 
Matt Hyclak
Department of Mathematics 
Department of Social Work
Ohio University
(740) 593-1263

Attachment: pgpxnRwbjEtUV.pgp
Description: PGP signature

<Prev in Thread] Current Thread [Next in Thread>