Review Trac backup procedure

[strk]

It turned out 3 trac instances are not being backed up, see ​https://lists.osgeo.org/pipermail/sac/2017-January/007874.html

This ticket is to review the procedure and ensure:

• All instances are backed up
• New instances are backed up (mostly implies documenting the correct procedure in ​https://wiki.osgeo.org/wiki/Trac_Instances#Creating_a_trac_instance)

[strk]

Wiki page updated: ​https://wiki.osgeo.org/index.php?title=Trac_Instances&diff=104712&oldid=104590 Backup script made to read /var/www/trac/trac_instances.txt and fixed the file to reflect reality. Now would like to change bacula config to use the script from its new location, but it isn't yet clear to me how does the director knows which file to transfer (as I see the backup.sh script prepares a .zip file but the path of the file I dont' see specified anywhere..)

[strk]

Figured: files to transfer are written in a FileSet section with Name="TracFile" found in bacula-dir.conf on the backup machine. Looking at it, there seems to be at least a useless directory (/osgeo/trac, which is now just 2 symlinks) and /var/www is missing, while it contains some statically written redirect pages. Any reason not to backup all of /var/www ?

Also, I guess we'd want another job (or extension in the "TracFile" FileSet ?) to backup Gogs data directories (as requested in #1657)

[strk]

Alright trac backup reviewed: we now have an /osgeo/backup/bacula_before.sh and an /osgeo/backup/bacula_after.sh scripts that are called to "prepare" the files to be backed up, and the whole of /osgeo/ is included in the backup, so the "bacula_before.sh" can take care of making hotcopies (which include database dumps) and putting them under /osgeo/backup, from which they are later removed by bacula_after.sh.

I've used the same layout for the osgeo6 machine, maybe can be standardized across all machines.

NOTE: /osgeo/backup on TracSVN is now under a local git repository