Re: [Duplicity-talk] Is it safe to modify a file when duplicity is backuping it?

[Kenneth Loafman]

Michael Terry wrote:
> On Sat, Mar 7, 2009 at 7:02 AM, Kenneth Loafman <address@hidden> wrote:
>> It's all about timing.  If the file is critical, you need to make sure
>> it stays in a cohesive state while duplicity backs it up.  If not, then
>> the next incremental will detect the changes.
> 
> It would be a nice feature to allow the user not to worry about this.
> What would be possible solutions?  Check 'modified date' before we
> start and after we're done?  If we detect a difference, doing the
> whole file over again seems bad.  Maybe just store an immediate
> incremental patch?

What we store now would be an incremental patch for the file.  If a
subsequent write modifies it again, then we would miss it on this
increment and catch it on the next.

The only problem is when an atomic change needs to be made to the file
that requires several writes (a database for example).  When we operate
in the middle of this, then we catch files in an intermediate state, so
 all backups are best handled on a quiescent system.

It's not really much of a problem, except for databases that are active,
and there's not one simple solution for all cases.  This is not limited
to duplicity, all backup systems have problems with an active system.

...Ken

signature.asc
Description: OpenPGP digital signature