[enhancement] Feature Request Email Notifications

[Alejandro Cuervo]

BIT stopped working about a week ago. I had not noticed, beacuse the lack of notifications of errors.
Email notifications would be great.

[Dan (danleweb)]

For now you can use user-callback script: http://backintime.le-web.org/documentation/usercallback/

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Julian Gilbey (jdg)]

This is not actually adequate for the purpose, because when the backup fails (see https://bugs.launchpad.net/backintime/+bug/1265499 ), the process exits without calling the callback with code 4. So I *still* didn't get notified that the backup had failed, even though I had the callback set up and waiting.

[Julian Gilbey (jdg)]

BTW, this means that this is not so much an enhancement request as a bug report.

[Julian Gilbey (jdg)]

Hello Germar,

I'm confused by the email notification I just received: what does "milestone none" mean?

I just got bitten by this bug again: for some reason, the backintime backup directory had ended up with wrong permissions (555 or 550, I don't remember exactly - no idea how it happened: nothing that I'm aware of having done), so every backup silently failed for about 3 weeks until I happened to check.

I've now written the following crude user-callback script to work around the problem; there is surely a better solution to this problem!

   Julian

#! /bin/bash

if [ -z "$USER" ]; then USER=root; fi

if [ $3 = 4 ] # error
then
    message="Backintime died with the following error information:
Profile id = $1
Profile name = $2
Error code = $4
Error message = $5

Just letting you know..."

    echo $message | mail -s "BACKINTIME FAILED" $USER
fi

if [ $3 = 2 ] # "Backup process ends" - this may work!
then
    dircount=$(ls -d /media/JG-BACKUP/${USER}/backintime/erdos/${USER}/1/`date ' +%Y%m%d'`* 2>/dev/null | wc -l)
    if [ $dircount -eq 0 ]
    then
        message="WARNING: Backintime finished, but no directory appears to have been created:
Profile id = $1
Profile name = $2

Just letting you know..."

        echo $message | mail -s "BACKINTIME PROBABLY FAILED" $USER
    else
        dirs=$(ls -d -1 /media/JG-BACKUP/${USER}/backintime/erdos/${USER}/1/`date '+%Y%m%d'`* 2>/dev/null)
        message="Backintime finished, and a directory appears to have been created:
Profile id = $1
Profile name = $2
Directory = $dirs

Just letting you know..."

        echo $message | mail -s "BACKINTIME finished" $USER
    fi
fi

exit 0

[Germar (germar)]

Hi Julian,

I'm not yet sure how to implement email notifications and after one year of working on 1.1.0 I really want to release that new major version now. So I skipped this enhancement for now (which means it will not make it into 1.1.0) but I'll pick this up again after the release.

Regards,
Germar

[Julian Gilbey (jdg)]

Hi Germar,

It's not even the email notifications which are the core of the problem - those can be wrapped around the backintime script. The heart of the problem is that when backintime fails catastrophically (no backup created at all, for example), it still exits with a zero exit code. This is just Bad.

In common/backintime.py, in the start_app function, we see code like:

        if arg == '--backup' or arg == '-b':
            take_snapshot( cfg, True )
            sys.exit(0)

Instead, if it were something like:

        if arg == '--backup' or arg == '-b':
            ret = take_snapshot( cfg, True )
            sys.exit(ret)

(or perhaps more sophisticated than this), and take_snapshot were tweaked to return a code indicating whether the snapshot succeeded or whether there were minor or serious errors, it would be automatic for cron to warn that there was a failed job.

HTH! Thanks,

   Julian

[Germar (germar)]

Hi Julian,

please take a look at version 1.1.0 released two days ago. This will do exactly what you suggested.

Regards,
Germar

[Julian Gilbey (jdg)]

Hi Germar,

That's great! Thank you! It will do most of the work, as now the cron job can detect whether the backup has succeeded or not.

   Julian

[Germar (germar)]

Quote from Mauro in bug #1434725:
Suggested options:
- send mail only when the snapshot fails or always
- attach snapshot log to the e-mail (only on failure or always)