Re: [rdiff-backup-users] incremental or differential backup

[Michael Biebl]

2009/1/16 Dominic <address@hidden>:
> Michael Biebl wrote:
>>
>> Hi,
>>
>> I'm new to this list, so if this question has already been answered
>> please bare with me and point me to the relevant discussion.
>>
>> I was wondering, if rdiff-backup stores the snapshots differentially
>> or incrementally
>>
>
> incrementally
>>
>> Say, I create a snapshot every day.
>> If I want to restore a file from 15 days ago, do I need all 15
>> snapshots to restore the state (incremental) or only the current
>> up-to-date state + 1 rdiff (differential)?
>>
>
> you need them all, but rdiff-backup handles it all 'under the hood'
>>
>> If rdiff-backup only allows incremental backups, I see the following
>> problems:
>> 1.) If a file changes a little every day (big mysql db), then
>> restoring the file (from say 100 days ago) will probably take a lot of
>> processing time, space and memory.
>>
>
> the extra overhead in space for many incremental backups vs. one
> differential backup is not great, but yes I guess it will take more time and
> memory

Does anyone have any first-hand experience with such a scenario (say a
several GB big mysql db, which basically changes every day)?

>>
>> 2.) If one of the rdiffs goes corrupt (e.g. via a bad sector), all my
>> older backups are broken.
>>
>
> hmmm, true I think, you should use raid or (better IMHO) a secondary backup
> (use rsync).

Is there a reason why you recommend rsync for that? If I understood
the project description correctly, rdiff-backup should work just fine
for remote backups.

Are there any (dis)advantages to rsync?

There is also the --verify-at-time option in rdiff-backup which
> I confess I have not used (oops!) - this should allow you (I think/hope) to
> confirm that backups at any given point in the past are not corrupt. Good
> idea to run this before doing the secondary backup I guess? [Question for
> expert: if rdiff-backup --verify-at-time 1Y succeeds does this also mean
> that all backups within the last year are uncorrupted too? I see that it
> reports success even if there are no backups that old in the archive.]
>>
>> 3.) I can't throw away rdiffs, say I want to create daily snapshots,
>> keep 30 of them, then keep monthly snapshots, and so on (basically
>> what tools like rsnaphost provide).
>>
>
> The logic of rdiff-backup is that you don't need to try to keep just monthly
> snapshots, you can keep your daily snapshots forever. It is true that
> recovering a file from a very long time ago, and which has changed a lot in
> the meantime, might take a while, but it would be a very rare event.
> Recovering more recent files which would be a more common event would be
> much faster, and of course the most recent backup (which is probably what is
> wanted in 95% of cases) is stored in the clear.

Again, does anyone have any first hand experience (and numbers) for
say a sever with a typical lamp stack (everything is backuped)

> Others may have some better informed comments...

Thanks for your comments so far.

Cheers,
Michael

-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?