Hello,
Daniel Holtkamp wrote:
Hi !
I`m playing with the bacula-configuration here (1.37.36) and i noticed
the following behaviour.
I started a backup-job to backup our Windows 2003 Server ... everything
ran fine till the windows-box crashed (still investigating why but this
is something else i think - not bacula related). Now the director says
the job is still running. Meanwhile i have added another backup-job to
the queue that is awaiting the end of the first job.
Now even if i cancel the windows-backup-job and the director says it is
properly canceled it still won`t go away ... from the documentation i
got the idea that it should continue after a few minutes but now i
waited an hour an nothing happened. (logfiles at bottom)
Well, I don't know which section in the manual you refer to, but this is
a network timeout problem and that's rather long in bacula - 2 hours, I
think.
So, you need to wait longer, and everything is ok. It works here, for
example.
Restarting the director solved the problem, alltough i had to start the
second job again.
That's normal.
The DIR doesn't store running but waiting jobs, so it doesn't know that
you wanted to run that job after the restart.
My concern is that when our bacula setup goes live and during the
nightly backup-run ONE client crashes during the backup and the complete
backup hangs ?
It just takes longer than you expected for bacula to consider the client
as dead.
Is there some sort of timeout that notices the client
isn`t sending data anymore ?
Yup.
Is there a way to continue the backup from
where it crashed (i doubt this functionality exists) ?
Right, that's not possible.
If i had a Full-Backup pool for the crashed machine with a volume
retention time of 2 months and 2 volumes maximum - what would happen if:
sunday fullbackup crashes - volume marked as used
monday - incremental gets upgraded to full - volume marked as used
tue-sat - incrementals as usual
sunday fullbackup -> retention time for both used volumes are not over
so what happens ?
I'm not sure I can follow your thought, but it seems that you found one
of the reasons why you always want to have as many generations of
backups as possible (possibility depending on available sorage, of
course) - the same problem arises when one set of backup volumes is damaged.
Bacula would notice a missing full backup when it starts a backup based
on a full backup - it needs the time of the last higher-level backup to
determine which files to save.
To my knowledge, though, there is one situation where it might miss the
fact that a full backup is not available: When an incremental backup is
dased on a differential one. Then, bacula would only notice the missing
full backup when the next differential job is started.
This is something Kern was thinking about or working on, though, so it
might be that bacula always makes sure that a full backup is available,
or this might become true in the future.
Arno
--
IT-Service Lehmann [EMAIL PROTECTED]
Arno Lehmann http://www.its-lehmann.de
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
