Re: [bitfolk] A question for users of the backup service - w…

Top Page
Author: alexvojtkoproctor
Date:  
To: users
Subject: Re: [bitfolk] A question for users of the backup service - what is a successful run?

Reply to this message
gpg: Signature made Mon Jan 11 20:59:50 2021 UTC
gpg: using RSA key 5C0FBC7994AA74932147BF1B5B5BD8731068A9C4
gpg: Can't check signature: No public key
In the situation that led to this discussion, I actually hadn't noticed
that some of the directories I was trying to back up weren't readable
by the backup user. Without notifications from monitoring, it would
probably have taken me a little longer to realise that not everything I
wanted was being included in my backups.

I don't see any harm in keeping things the way they are. There's an
easy enough way of excluding potentially unreadable files from backup
if you definitely don't want them included.

On 2021-01-11 15:57, Andy Smith wrote:
> Hi,
>
> A question for those who use the backup service¹:
>
> Currently we mark the backup run as successful if rsync exits with a
> success value.
>
> There are only two exceptions: exit code 2 and exit code 24. Both of
> those relate to files which rsync thought existed but ended up not
> existing when it came to actually transfer them. I consider those
> transient issues related to backing up a filesystem that is in use,
> and not a reason to consider the whole backup run as failed.
>
> So what about files that our rsync process cannot read? At the
> moment that produces an exit code of 23, and is considered a failed
> run, even though everything else got transferred. This eventually
> causes a "backup age" monitoring alert because the last successful
> backup run was too long ago. Even though everything else is actually
> being backed up.
>
> If we consider error code 23 as okay then a backup run that failed
> to transfer one or more files due to permissions is still considered
> a success and the alert goes away. But you possibly never find out
> about what happened because you don't get to see the logs, you would
> have to check every file in your backups to be sure they're there.
>
> If we continue to consider error code 23 as a failure of the whole
> run then you will have to either allow our rsync to read the files
> concerned or else put up with perpetual alerts - which you could
> silence but then would never tell you about other problems.
>
> What should we do?
>
> Note that most of you allow our rsync to run as root so it can
> generally read everything and you'll never experience this. But in
> theory you could if you found some way to deny root permission to
> read something.
>
> I would ask for opinions only from those who make use of the backup
> service, as root or not.
>
> Cheers,
> Andy
>
> ¹ https://tools.bitfolk.com/wiki/Backups
>
> --
> https://bitfolk.com/ -- No-nonsense VPS hosting
>