I am in the process of choosing and evaluating between 2 free network backup
software: amanda and bacula.

The tape device is a DLT8000 that is in a StorEdge L7 tape library.
It is attached to an old Sun E220R running Solaris9.
mtx has been installed and is working.
Client hosts to backup are SPARC machines running Solaris8/9 and x86 boxes
running Linux.

Has anyone tried bacula on Solaris ? ... and how was your experience with
it ?

I was initially consdering amanda ... but it seems that amanda is very
complex compared to that of bacula.

Appreciate any comments.

jms <noone@noone.org> writes:


>I am in the process of choosing and evaluating between 2 free network backup
>software: amanda and bacula.

>The tape device is a DLT8000 that is in a StorEdge L7 tape library.
>It is attached to an old Sun E220R running Solaris9.
>mtx has been installed and is working.
>Client hosts to backup are SPARC machines running Solaris8/9 and x86 boxes
>running Linux.

>Has anyone tried bacula on Solaris ? ... and how was your experience with
>it ?

Sure. We're running it for the computer club here and has so been
doing that for soon a year now.

Backup server:
	Sun Ultra 60, 2x360MHz, 2GB RAM, 6 SCSI controllers
	Solaris 10
Tape robot:
	Sun StorEdge L11000 (6 DLT7000 drives, 384 slots)
Backup clients:
	A number of Suns and PC's running various variants
	of Solaris and Linux. Some small, some large ($HOME
	and mail servers (maildir -> huge amount of files).

It works mostly well and is getting better and better every day.

The main problems we've seen is that it's not really optimized for
*big* fat servers with many files - but then I don't know of any
backup programs that handles that situation really well (not even
the commercial ones like Legato Networker that we use at my department).

The problem here is that it takes a *long* time for Bacula to generate
a "view" of a large filesystem when doing restores (it builds a view of
the whole file systems in RAM). Be prepared to have much RAM on your
backup server (we just upgraded from 1.5GB to 2GB RAM due to problems
with an attempted full recover of one of the large servers). A fast CPU
is probably also advantageous.

On the good side thought is the fact that it *does* work with big and
large filesystems and handles multiple backup clients - and uses
standard SQL databases for the indexes. It also has some other good
features like being able to stage backups from clients via spooling files
(decreases the backup window on the clients and allows the tape drives
to stream) on the server, being able to have the index (SQL) databases
on one system, the controlling daemons on another, the tape drive
controlling daemons on any number of systems. It also supports backup
to files or CD/DVD and not just tapes.


By the way, "big" filesystems is those that have a lot of files - the
size of the files isn't really relevant. Our $HOME has around 2-3 million
files, the MAIL server has over 4 million tiny itsy-bitsy files...

I'd definitely recommend that you check Bacula out and test it.

- Peter
-- 
-- 
Peter Eriksson <peter@ifm.liu.se>            Phone:    +46 13  28 2786
Computer Systems Manager/BOFH                Cell/GSM: +46 705 18 2786
Physics Department, Link�ping University     Room:     Building F, F203

Thanks for feedback. A few questions though:

Peter Eriksson wrote:

> 
> The problem here is that it takes a *long* time for Bacula to generate
> a "view" of a large filesystem when doing restores (it builds a view of
> the whole file systems in RAM). Be prepared to have much RAM on your
> backup server (we just upgraded from 1.5GB to 2GB RAM due to problems
> with an attempted full recover of one of the large servers). A fast CPU
> is probably also advantageous.

Does that mean you need to have RAM or virtual memory that is as big as the
biggest filesystem that you want to backup ?

> 
> On the good side thought is the fact that it *does* work with big and
> large filesystems and handles multiple backup clients - and uses
> standard SQL databases for the indexes. It also has some other good
> features like being able to stage backups from clients via spooling files
> (decreases the backup window on the clients and allows the tape drives
> to stream) on the server, 

That is also what I needed. I know amanda does streaming to tape by backing
up clients hosts to the filesystem on the backup server first ... and then
streaming the files to tape. However, I could not confirm this with the
Bacula documentation ( I have not read all the docs, however ).

One final question ....

Can you restore from tape even if the bacula software is not installed ?
e.g. ... You bring in a bacula-labeled tape to another location and want to
restore some specific files, or the backup server was replaced with a
newere machine and bacula have to be reinstalled.

jms <noone@noone.org> writes:

>Thanks for feedback. A few questions though:

>Peter Eriksson wrote:

>> 
>> The problem here is that it takes a *long* time for Bacula to generate
>> a "view" of a large filesystem when doing restores (it builds a view of
>> the whole file systems in RAM). Be prepared to have much RAM on your
>> backup server (we just upgraded from 1.5GB to 2GB RAM due to problems
>> with an attempted full recover of one of the large servers). A fast CPU
>> is probably also advantageous.

>Does that mean you need to have RAM or virtual memory that is as big as the
>biggest filesystem that you want to backup ?

No - but you must have enough RAM or VM so that the Bacula daemon (one
of them - the Director) can build up a BTree containing all the filenames
(and some other information, like modification times, file sizes and which
tapes (and where on the tapes) the files are stored on).


>> On the good side thought is the fact that it *does* work with big and
>> large filesystems and handles multiple backup clients - and uses
>> standard SQL databases for the indexes. It also has some other good
>> features like being able to stage backups from clients via spooling files
>> (decreases the backup window on the clients and allows the tape drives
>> to stream) on the server, 

>That is also what I needed. I know amanda does streaming to tape by backing
>up clients hosts to the filesystem on the backup server first ... and then
>streaming the files to tape. However, I could not confirm this with the
>Bacula documentation ( I have not read all the docs, however ).

Bacula uses "spool" files on the tape drive controlling server to
stage the backups - these spool files you typically limit in size
so they don't use up all available disk space. The spool files typically
are smaller than the file systems you backup.

Our setup uses one 9GB SCSI disk per spool file (one spool file per
tape drive). Like this:

/var/mysql	- 100GB (index database - mirrored 36+73GB disks)
/var/bacula	- 18GB (various files, mirrored 18GB disks)
/spool/dlt-0	- 9GB (spool file for DLT drive 0)
/spool/dlt-1	- 9GB (spool file for DLT drive 1)
....


>One final question ....

>Can you restore from tape even if the bacula software is not installed ?
>e.g. ... You bring in a bacula-labeled tape to another location and want to
>restore some specific files, or the backup server was replaced with a
>newere machine and bacula have to be reinstalled.

I believe there are "Live CDs" available (bootable Linux ones with Bacula),b
but I have never tested those. Getting Bacula running is quite simple so
I haven't really bothered.

- Peter


-- 
-- 
Peter Eriksson <peter@ifm.liu.se>            Phone:    +46 13  28 2786
Computer Systems Manager/BOFH                Cell/GSM: +46 705 18 2786
Physics Department, Link�ping University     Room:     Building F, F203