On Wed, Mar 15, 2006 at 06:05:05PM +0100, Thomas Widhalm wrote:
> Am Mittwoch, 15. März 2006 17:27 schrieb Jon LaBadie:
> 
> > I know you are not using tape now, but whenever you run
> > amcheck (you do don't right?) it will complain tapecycle
> > less than runspercycle.
> 
> I do call amcheck from a script (just su's to amanda and then runs amcheck. 
> It 
> doesn't complain.

Maybe you are running with some option like -c which eliminates
tape system checks.  If not, then your amcheck has serious probs.

> > The general way to reduce full dump frequency is increase dumpcycle.
> 
> I tried this. But it just overloads my harddisks,
> because there are so many level 0s.

Doesn't make sense to me.  Currently you say do a full dump
every 8 days.  To change it to every 30 days should NOT
cause more overloading than 8.

> >
> > I don't think this is the problem, but I do not see a "record yes"
> > setting.  I wonder if that is needed to ensure amanda knows when
> > the last full dump was done?  It is needed for "dump", but I'm
> > uncertain about gnutar.

When I grep'ped your original config posting I found only one
instance of "record".  It is in this dumptype:


  define dumptype nocomp-test {
    global
    comment "test dump without compression, no /etc/dumpdates recording"
    compress none
    record no
    priority medium
  }

Is that a dumptype you use or include into one you do use?

Another way to check the setting is with amadmin.  The disklist
command prints out many of the parameters for each, or the
specifiec, DLEs.  See if you get record yes or no.

> 
> The docs say, record has default value "yes".
> I tried to include it, but amcheck complains,
> that it's no valid option.

Then type it correctly :)  And in the correct location.

> 
> Maybe I found it deep within the  comments. There is one line "reserve 0". I 
> can remember I changed this on the very day I got to use this server, for it 
> didn't backup anymore. Or maybe it was some similar option. The comments 
> read, that this just applies to backing up to the holding disk and "reserve 
> 100" means, that there are just incremental backups and that a smaller value 
> gets amanda to do fullbackups till the holdingdisk is full. So I tried now 
> and set it to "reserve 50". We are very well beyond that now so amanda should 
> just do incremental backups by now. I can imagine that when it comes to 
> deleting old backups there will be enough free space for 1 or 2 level 0 for 
> every disk. I will test and watch because this could mean that sometimes I 
> miss a level 0 for a disk. Maybe a mixture of "reserver 50" and a cronjob 
> doing a forced full every tapecycle like somebody mentioned, will do the job.
> 
> I will report the outcome of this.

I look forward to it.  But I fear it will only mask an underlying
problem.  Up to the reserve you are correct, amanda can put level 0
backups onto the holding disk.  But it can also put incrementals
there as well.  I.e. up to the reserve, it can do its normal
scheduling.  After that, things scheduled for level 0 backups
are "demoted" to incrementals.

The actual question is why is amanda scheduling so many level 0s.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)