> >  > The job for web22 for example has been running for 2 days.  Most of
> >  > the other jobs finish in about 16 hours but there are some stragglers
> >  > with lots of data that take longer.  To prevent this I skipped an
> >  > incremental the day after the full but since it's been 2 days, the
> >  > next incremental kicked in and was upgraded to a full.
> >
> >  Yes, this is a known issue - not a bug, but definitely something that
> >  can be improved.

Kern argues it's a feature. I disagree

> >  > In the past this just caused the second backup to be put on hold until
> >  > the first finished.
> >
> >  Sure? I don't think I ever observed this.

I haven't either. It's been a fundamental annoyance for me from the outset
of using Bacula a few years ago.

> One way you could get around this is to make it so only 1 job can run
> at a time for a particular client. This will allow the full backup to
> complete and the incremental queued to not actually start until such
> job is completed.

Doesn't work. The incremental notices the full hasn't completed, upgrades
itself to full and THEN queues itself.

IMHO this order is incorrect. The check should be made as the process
exits the queue and starts running.

> But I'd have to agree, what are you backing up that is taking more
> than two days? I hope you say you've got terabytes of data streaming
> along or there is something else you should be worried about.

1Tb takes about 26 hours here due to spooling time. 2 * 1Tb backups also
takes about 26 hours...

This is just enough to tickle the behaviour.

AB

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Register now and save $200. Hurry, offer ends at 11:59 p.m., 
Monday, April 7! Use priority code J8TLD2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users