Re: [Bacula-users] configuration advice (usb disks and filesets)

[Arno L. <al...@it...>]

Hello,

On 05.10.2005 23:19, David Catanach wrote:

> Please forgive me if something like my situation is covered in a
> previous post or in the instructions somewhere. I found one similar
> question that no one had answered... here's hoping someone does.

Nothing to forgive until here - at least you looked before you posted!

> I'd like some advice as to how to set my filesets up. We have three
> machines backing up presently: one is the bacula director/storage host
> (approx 100G over 12 partitions) and two windows hosts (probably around
> 20G each). We will be backing up to disk (more about that later) Would
> it be advantageous to configure a fileset for each partition on the
> linux machine? I would think this would speed up restores as each tape
> image would be shorter. However this complicates the scheduling because
> I'd need a seperate job for each fileset, correct? Is there be any way
> to group jobs, sort of aliasing them together under one name, like per
> machine?

The latter question - No. Unfortunately.
"Would it be advantageous...?" - You answerde it yourself, I think: I 
complicates scheduling, and, in my opinion, a backup schema has to be 
one thing first and above all: Simple. You don't need errors in it, and 
in case of emergency, others have to understand it easily.

> The other bit of advice I'd like to ask for has to do with pool/volume
> setup. We want full backups (monthly) to go directly onto a USB-attached
> hard drive, along with the catalog for that backup for offsite storage.

I hope you did consider the reliability of external hard disks, 
especially when they are moved around. (If you really want to use 
external disks, use 2.5" drives for notebooks, and operate them with a 
proper power supply.)

> Daily incrementals will be written onto a directory on the linux
> machine, along with weekly differentials. The differentials may also be
> written out to cd/dvd (manualy or automaticaly after spooling, depending
> on how long it takes me to sort the rest of this out, not so important
> for the question)

Hrm. Sorry to hear that, I'd be very interested to hear more about DVD 
writing with bacula ;-)

> The full backups would be rotated among N (N = 4
> currently) USB disks which would be recycled after N months to provide N
> months of weekly backup + daily granularity depending on how often the
> incrementals are pruned.
> 
> Whew.

Indeed.

> I can't determine if I need one pool for full backups, or N pools for
> full backups, or N * hosts pools for full backups, or if it really
> matters.

Well, it matters in so far that your setup must reflect your considerations.

> I could just mount a new disk each week, but then I have to
> keep track of which one I'm using manualy, so mistakes will be made.

I would do the following, but keep in mind that I never set this up:
- Use one storage device definition for all your usb disks.
- Make sure your OS mounts all the drives to the same mount point (not 
exacly easy with automounting by hotplug and/or HAL etc.)
- Put big stickers onto all your hard disks: "ONLY ONE OF THESE DRIVES 
MUST BE CONNECTED TO NAME-OF-BACKUP-SERVER AT ANY TIME. NO EXCUSES!"
- If you've got you disks partitioned and formatted, set up the pool 
with a maximum number of volumes reflecting your available disks, set 
the maximum volume size according to your disks, set the retention times 
according to your needs, and label the disks one after the other. Don't 
forget to unmount the disks before removing them.
- Make sure bacula accepts any suitable volume - either set that 
directive explicitly, or use the development version.
- Install a Run Before Script for all jobs going to the disks that 
checks if a disk is mounted, and if it's not, let the job fail, or send 
mail, wait, check again ad infinitum.

I imagine that several things can happen now: Bacula finds the volume it 
wants. It uses it, everythings's fine.
It finds a wrong volume, and it either ends the jobs in error (you know 
what has happened, at least) or it takes whatever volume it can use (as 
long as the volume is in state Append or Pruned). This would be best, of 
course.
If it fills a volume, it should ask for another one.

Try it, and we all might learn if bacula asks for user intervention even 
with disk based volumes :-)

> If
> bacula were writing to a real tape device, I would think to use N full
> pools (one for each set of tapes to rotate on),

Nah, I'd rather use one pool for Full backups and set retention times 
and, probably, use duration according to my needs.

> but is there any way to
> have bacula know which USB disk I have plugged in, or which I should use
> for a restore? Can bacula only track by 'tape' label for disk storage?

It should, as far as I know, but I never tried it.

> Also, can I have different pools point at different 'devices' or
> locations on disk, so that fulls go onto the usb disk and diff and incrs
> go onto some other filesystem?

Yes, but you need to set up a storage device for each of these disk 
directories. To easier management, you can use the great invention of 
symlinks when restoring...

> Or maybe I'm thinking about this all
> wrong and someone can set me straight.

I think there are many possible solutions, but it's your decision.

> All my configs are available at www.nmt.edu/~dcatanac/bacula/
> 
> Thanks a bunch in advance.
> 


Arno

> 
> -------------------------------------------------------
> This SF.Net email is sponsored by:
> Power Architecture Resource Center: Free content, downloads, discussions,
> and more. http://solutions.newsforge.com/ibmarch.tmpl
> _______________________________________________
> Bacula-users mailing list
> Bac...@li...
> https://lists.sourceforge.net/lists/listinfo/bacula-users
> 

-- 
IT-Service Lehmann                    al...@it...
Arno Lehmann                  http://www.its-lehmann.de