OK, to indicate where the problem is..
I have these assumptions about backup-jobs (tell me if some of them are
insensible):
1) Full and differential backups must always run on friday - otherwise jobs
lasting more than 12 hours may disturb working.
2) If full or differential backup fails, it must be ran again - otherwise full
backups may start aging and thus needed backup SLAs may not be met.
3) Duplicate jobs must not run - if a full job lasts more than 24 hours and
there is another job starting in 24 hours, full backups will cycle forever.
Anyone else having these assumptions?
To my understanding it is currently not possible to configure this kind of job
with options ("Rerun Failed Levels = yes" and "Allow Higher Duplicates = no")
meant straight for that, because they do not work well together.
The best compromise seems to be throwing out assumption 2 and setting "Max
Full Interval" to force full backups being up-to-date, but this setting may
cause a full backup to start on a wrong week-day (that could be a serious
problem in some cases).
--
Silver
On Thursday 25 March 2010 09:26:05 Silver Salonen wrote:
> On Monday 22 March 2010 11:26:04 Silver Salonen wrote:
> > Hello.
> >
> > I'm using directive "Allow Higher Duplicates = no" in 5.0, but I find its
> > behaviour a bit false in case "Rerun Failed Levels = yes" is set -
cancelled
> > duplicate job makes the same level to be ran again. The problematic in
this
> > behaviour is this use-case:
> > 1) a full job is ran
> > 2) another full job is ran - it's cancelled as duplicate
> > 3) the 1st full job finishes successfully
> > 4) if you run incremental now, it will be upgraded to full, because of the
> > cancelled duplicate job
> >
> > The problem is reported in http://bugs.bacula.org/view.php?id=1507, but as
> > Kern explains there, it's not so trivial fixing this.
> >
> > Maybe the community has any good ideas for this?
> >
> > PS. Using "Max Full Interval" with "Rerun Failed Levels = no" could help
> > somewhat, but I guess it doesn't fix the same loop-hole for level
> > differential, ie. it could still mess up a bit of GFS rotation.
>
> Anyone interested in the problem?
>
> --
> Silver
------------------------------------------------------------------------------
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users
|