Do snapshots solve all consistency problems ?

[scdbackup@gmx.net]

Hi,

this thread was originally named
"mkisofs aborts but exit value is 0"
but went into the fundamental question wether a
backup will allow a sufficient restore.
Sufficient for the needs of the person or organisation
which created the backup. So that they can go on
with their operations as if there was neither a
restore nor a disaster before the restore.
Only the system history between backup and disaster
recovery may be lost.

I acknowledge Joerg's effort to make this as safe
as possible by help of his program star. I consider
to learn about LVM in order to get filesystem 
snapshots.
Nevertheless i do not share his optimism about the
universal usability of a properly recorded snapshot
backup.


>> I wrote:
>> Does a snapshot really decrease the probability
>> of backing up a file while it is in an ill state ?
>> "Ill" in the sense that it does not comply to a
>> valid persistent state as expected by its applications
>> when they open it.
>> Not "ill" in the sense of "a case for fsck".
> Joerg Schilling wrote: 
> If you have an ill designed application, you are lost.
> decent applications should write files in a way that
> does not create ill states.

Shall i really make assumptions about the quality
of particular applications while i consider
potential backup problems ?


Meanwhile i have the idea that shapshots decrease
the probability for unusable file copies in the
same way as quickly crossing a street reduces the chance
to collide with a car. Slow people get overrun more
often. But you can also run blindy into a standing car.

This, of course, only if there are any cars on
the street.
If we put all cars into their garages, then
there will be no accident. Regardless of speed.
Runlevel 0.

Another approach would be to learn how to detect cars
and to wait until the situation is safe.
(But if all cars stand still, a hampering car
 would never drive away and thus block the crossing
 forever.)


>> A snapshot does not keep you safe if you don't know
>> exactly what particular services need to be shutdown
>> temporarily.
> 
> When using snapshots, you don't need to shutdown services
> while running a backup.

I still do not see how a snapshot does prevent
to snap intermediate file states which do not
suit a service when it opens the file.
After restoring a backup such a service or application
would make problems.

Trying to imagine how i would have to redesign the
applications on which i worked in the past ...
... so that all file changes happen atomically ... 
... i encounter severe obstacles.


Are there guidelines on how to achieve this quality
within general purpose programs ?

A sysadmin would need a way to determine
wether his system components comply. Is there some
directory where to learn about snapshot safe
software ? (I would like to learn their tricks.)


Have a nice day :)

Thomas