> That would be a harsh rule to enforce.  

I think that depends on which you consider more important: (a) to back
up tonight, or (b) to ensure that yesterday's state is re-achievable
in the event of disaster even if that means losing today's work. I
think (b).

> Suppose my small company
> backs up 7 desktop computers.  If one DLE on the next tape was
> the last level 0, what should amanda do during that amdump?
> It can't "tape around" the problem dump file.

I guess you think (a). Fair enough.

> Should it skip the dump leaving lots of new stuff not backed up?
> 
> Go into degraded mode and hope the holding disk will hold all
> the forced incrementals?
> 
> Copy the file from the tape to some archive area, hopefully
> with sufficient space.

Either of the first two (plus a "your backups failed" email) would
be fine by me. The third option sounds like a second (v)tape library,
which, if we had one, we'd allocate in the first place.

> IIRC, amcheck does give warnings about the impending situation.
> I'd say it is up to the admin to decide how to handle it.

Ahh ... good! Still, it would be awful to come back from a long
holiday, and discover that, despite those warnings, the only full
backup *had* been overwritten and that the system itself had then
died of disk failure.

I guess the only option is to have a tapecycle big enough to cover
all eventualities, as the Tapecyle wiki page suggests.

Thanks for the feedback!

Alexis

Reply via email to