>
> I may be soon setting up a "disk swapping" system, so would be interested to
> see how you managed it.
>
> j
>
> --
> Joshua Kugler
Ok Joshua;
The backup machine:
AMD64 dual with 4 Gigs RAM
Gigabit ethernet on a gigabit switch
a 80G IDE just for freeBSD 6.1 / Mysql 5
a 120G dirve fo bacula 1.38.9 (bin, working and working.drive1, etc...) mounted
on /bacula (owned
by it)
a 120G just for the database mounted on /db (owned by mysql
a removable tray for the 400G SATA, the backup drive mounted on /bck (owned by
bacula also)
I didn't compile from the ports. Downloaded from bacula.org and compiled
directly.
under /bck, I created a directory for each host, with the exact name of the
clients on bacula-
dir.conf.
Due to the cranky nature of our mysql implentation, I hardcoded
DB Port = 3306
DB Address = 127.0.0.1
DB Socket = /db/mysql.sock
inside src/cats/mysql.c. I noticed that the db_connect function doesn't pass
down neither one of
these. It was as if it was not using those conf parameters so I forced them in.
That way, all three deamons will use it. Before doing that, I could not get
Director to connect to
mysql, although I had set up the above inside the Catalog resource. Well, real
tiny problem.
1) I set up the schedules, monday being the full dump day, for 9 of the servers
aand an hourly
backup for the database server.
2) created 10 restore jobs to restore any data to any machine I want, case I
need it.
3) Created separate Client, Job, Storage, Pool and FileSet in bacula-dir.conf
and a Device pointing
to /bck/hostmane in bacula-sd.conf, FOR EACH OF THE HOSTS ( Catalog and the
backup server itself
included).
Since bacula does a great job in labeling each file already, I wrote a simple
set of scripts that
keep track of the harddisks themselves, their names, dates of insertion, dates
it came out and when
it was used to restore data.
I don't let bacula erase bacula.sql, so it remais inside the working/ dir,
besides being backed up
with everything else. When a disk is swapped, the current working dir is moved
to, for instance,
working.disk-01.22-06-2006.08:00, a new and empty working/ is created for
disk-02 that just
came in and became current. If I need to restore data for disk-01, working is
preserved by
becoming working.current, working.disk-01.22-06-2006.08:00 (with its own
bacula.sql !) becomes
working, the database is cleared and loaded with the backup.sql that is there.
Pronto! everything
is back to what it was when disk-01 was last seen there, with the exception
that all backup jobs
are disabled during the restore session. Reverse the process to get the current
disk back when done
restoring.
The scripts do all that, keep track of what disk is current and won't let you
bring back the wrong
disk.
There are 4 scripts:
hdstatus - Tells - which disk is in there and what date/time it was put there
- when each disk was in/out of the sistem (Insertion
history)
- how many restore sessions where ran, with which disks
and when
(restore history)
- how much each host is occupying in its /bck/hostname
directory and if
- a backup is running on any of them
- which HD's working.disk-nn.* are present
hdswap - Does what it says
hdrestore - prepares the system for a restore session
hdreturn - To return the current disk to the system.
On every root login, hdstatus is executed to tell you if you forgot to return
the current disk
after a restore session, and to remind you that backups are disabled until the
current disk is
returned.
Well, that the general picture. Of course there are many other implentation
details.
The machines being backed up are RH8, RH9, Fedoras, FreeBSDs 4/5/6 and windows
2000 and 2003.
Except for the windows clients of course, all the clients were compiled on the
servers themselves.
Sorry for writing so much.
I hope it helped
--
//| //||
// | // ||
-//--//--|| ARIO LOBO
// // ||
---------------------------------
[EMAIL PROTECTED]
http://www.digiart.art.br/MarioLobo/
http://www.mallavoodoo.com.br
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
