Hello,
On 3/21/2006 9:39 PM, John Goerzen wrote:
Hello,
I have been using Amanda for backup for quite a few years now. I'm
interested in Bacula and have read through most of the (large!) manual.
Bacula looks like a nice piece of software, but there are several things
that concern me about it.
1) In the manual, it states that "if you move files into an existing
directory or move a while directory into the backup fileset after a Full
backup, those files will probably not be backed up by an Incremental."
That's concerning to me and seems a huge hole -- files are moved or
renamed all the time in our environment, and we would want them to still
be backed up. Some other places in the manual, it suggests to copy
instead of move/rename directories. But with a user base of several
hundred people, some of which aren't all that computer-savvy, that is
almost impossible. In the Win32 area, it had another warning about the
same problem, with no apparent workaround other than to copy it. Again,
a serious problem. It seems that, as a temporary solution anyway, it
would be easy enough to just treat files inside moved/renamed
directories as new and back them up.
The temporary solution you suggest is part of what would be needed to
implement a final soultion :-)
The problem there is not a conceptual one, it's rather that nobody
implemented it yet. As you note, all the necessary data is available,
it's only necessary to modify the FD to check each file against the
catalog database, and for that purpose, I think, the DIR-FD-protocol
needs to be modified. I think this is a rather large project.
2) I'm concerned that incremental and differential backups don't notice
deleted files. When we restore from that, we could wind up with
thousands of deleted or renamed files -- *not* an exact image of the
system as of the last backup. That also is a large problem.
See above...
3) We perform backups overnight, when no operators are here, so as to
minimize performance impact on our users. We have enough data that it
is not possible to fit a full backup of every filesystem onto a single
tape. However, it is possible to stagger the full backups so that we
guarantee each filesystem is backed up with a full backup once every
three days. Amanda will automatically handle that staggering for us.
Can Bacula?
Not automatically. Instead, Bacula can use spooling, it can use
autochangers, and it allows you to set up schedules as you want them. In
the future, Bacula will allow job migration, opening up the whole new
world of D2D2T backups.
4) Amanda has a nice "degraded mode". If the tape drive is offline, or
there is no tape in the drive, or the tape in the drive is not suitable
for backup, Amanda will run all incremental backups and store them to
disk. The next morning when the operators arrive, they can correct the
problem and run amflush to move the data out to tape. I notice that
Bacula has the ability to spool data to disk, but it doesn't appear to
be able to do that in a nice fashion in the presence of a tape error.
Correct?
Depends. Usually, tape errors result in the request for a new tape, and
won't break running jobs. Once all data is spooled, it will wait until a
usable tape is mounted and then continue despooling.
5) More generally, I am concerned about this notion of continuing to
append to a tape until it is full. We would not know in advance when a
tape will fill up. Simply waiting for the operator to swap tapes, and
then continuing with the backup, is not a workable solution for
performance reasons since we can't run backups during the day. We would
be left with missing a night's backup. Is there any better way we could
use Bacula's nifty append features in our setup?
Well, there are quite a number of ways to automatically mark a tape as
Used, i.e. no longer appendable. Which one you chose depends on your needs.
6) And even more generally, are people actually using Bacula in medium
to large organizations?
There are setups I consider large. I know of some multi-TB setups, and I
think you can find reports in the mailing lists about the problems
native with such setups, like backup runs that take days and catalogs
that grow beyond what the database handles without tweaking.
Reading about how the author's test network is
using token ring, how some sites only have to swap tapes once a month,
and how many are able to keep using the same tape day after day makes me
think that Bacula may not really be suited for a situation in which we
store terabytes of information and back up dozens of machines. Is that
an incorrect assumption?
I'd say so... swapping tapes only once a month or reusing the same tape
over and over should, IMO, only be mentioned in the section "Things to
avoid if you want a Backup" :-)
Arno
Thanks for all your work on Bacula and its manual.
-- John Goerzen
-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
--
IT-Service Lehmann [EMAIL PROTECTED]
Arno Lehmann http://www.its-lehmann.de
-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
