Hi there,

I've wanted to chime in here before now, but I thought it might all
blow over.  It didn't.  I want to try to inject a dose of reality.  I
have no axe to grind over BackupPC, and, except that maybe I see some
unrealistic challeneges, I'm not trying to defend it.  I'm a satisfied
user, that's all.  Sometimes I pipe up here on the list to try to help.
Sometimes I've said things not unlike what I'm going to say below.  It
is intended to help but it might seem a bit harsh.  I apologize now.

On Wed, 21 Aug 2024, Ghislain Adnet wrote:
On Wed, 21 Aug 2024, Alexander Kobel wrote:

...
> If I understand correctly, the situation here is:
>
> 1) Backups have been executed in the past and been working fine.
> 2) Some corruption on the server happened; e.g., one or several
>    pool files experienced bit rot.

I take issue with the term 'bit rot'.  Either a filesystem works - and
in that case there is nothing, ever, which you could (mis)classify as
'bit rot' - or it doesn't, in which case it's *broken* and should, if
anyone relies on it for anything, be taken out of service either to be
fixed or to be replaced.  In my view there's no middle ground.  If you
do not reliably get out exactly what you put in then it's just broken.
There will occasionally be 'soft' damage which can be repaired.  There
are various ways of doing that with various degrees of confidence that
the finished repair is 100% good.  I don't trust any of them.  Far too
often I've seen what looks like a repaired filesystem throw a spanner
in the works when it turns out that it wasn't really repaired at all.

> 3) BackupPC_fsck detects the issue and reports a file as broken.
> 4) ... but there is no means for BackupPC itself to *react* on
>    the issue; it simply warns, but there is no recovery.

BackupPC's main job is to copy files from Known Good FileSystem A to
Known Good FileSystem B on the basis of Known Good Metadata.  It does
what the docs say it does.  I have many years of experience with it,
so I'm confident of those statements.

If you have a problem with one of your filesystems, all bets are off.
Strike out the "Known Good" parts in the previous paragraph, and what
you have left is nothing that's going to be very much use to anybody.

BackupPC is not a fault-tolerant storage system and it never claimed
to be.  For that you're looking at things like RAID systems.  Money.

If a filesystem is damaged you have to fix it yourself; maybe BackupPC
can help, to the extent that it *might* be able to provide copies of
data from the filesystem.  But if whatever broke it did a thorough job
of the damage, then it's probably time to cut losses and start again.

Potential issues include crashed heads; marginal logic levels; dodgy
applications; components sensitive to temperature; moisture attacking
connectors; even something as simple as the lack of reliable power.
Even if software *did* manage to repair it, there'd be no way to know
that it would last the next hour - never mind the next backup cycle.
It's a minefield.  There's a whole industry devoted to dealing with it
and to suggest that *any* one piece of software might be able to cope
with that is to show that you have no appreciation of what's involved.

 ?I can have issue but then i need to be able to

- spot them

Apparently you've spotted an issue.

- correct them

in a programatic way (humans are unreliable).

That's like saying your car shouldn't stop when all the coolant leaks
out of the radiator.  You had a temperature gauge.  You were supposed
to keep an eye on it.  You failed to do that.  Now, instead of a can
of Radweld, you need a new engine.  It's entirely your own fault.

You seem to be asking BackupPC to be a filesystem repair tool.  That's
a completely different animal, way beyond the scope of a comparatively
simple scheme for copying files, and unless the latest Pacific Bubble
'AI'[*] turns out to be more than a bubble there never will be any way
in general, to repair a broken filesystem reliably and automatically.

I've repaired many broken filesystems in my time.  Anything from trips
over the power cord to fire caused by a cat asleep on the monitor. [**]
Early on, one of the things which this taught me is that the number of
ways in which ONE filesystem can be broken approaches infinity - I have
never seen the same breakage twice - never mind the fact that there are
at least a dozen filesystem *families* in current use.  As you'd expect
BackupPC works on nearly all of them, but it makes no sense to expect
it to be able to repair them too.

It is just not possible with the current state of the art.  About the
best we can do is something like fsck (not BackupPC_fsck, but the real
thing that comes with your Linux box) and even that (a) only handles a
limited range of damage to a limited range of filesystems (b) can, and
often will, give up on the job and ask for your help and (c) can never
guarantee that there isn't still something left in there that's broken
- for that you'll need levels of integrity checking which cause levels
of aggravation far exceeding the average computer owner's tolerance.

[*] I think that by the time 'AI' matures enough to be really useful
there probably won't be any need for anthing like what we presently
call a filesystem.

[**] I kid you not, in the milking parlour at a Nottinghamshire farm.

Backuppc_fsck takes more than 24h on the backuppc server so cannot
be used "regulary" without preventing backuppc normal operations.

At the very worst you can schedule around it.

anyway the issue is here to spot and correct problems. If there is a
way, be it? in the rsync options or backuppc system i would gladly
hear it as perhaps i missed something somewhere :)

If I have the slightest suspicion that I have a corrupted filesystem I
will replace it immediately.  Quicker if it's used by a backup system.

There's no point trying to limp along, because sooner or later it will
come back to bite you - most probably (1) when you least expect it and
(2) when it will cause maximum inconvenience.

If you do not maintain your filesystems in top condition then you are
never going to be very far from a disaster.

You just won't know how far until it happens.

--

73,
Ged.


_______________________________________________
BackupPC-users mailing list
BackupPC-users@lists.sourceforge.net
List:    https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki:    https://github.com/backuppc/backuppc/wiki
Project: https://backuppc.github.io/backuppc/

Reply via email to