Simplify and make program calls more consistent with other handlers. The
automatic retry mechanism shouldn't be needed for restic any more than
other backup programs. If it's implemented, it should be at the core
level.
When testing the connection via ssh forced command, if the remote path
is an empty directory (instead of nonexistent), the handler bails even
tough the connection is successful.
Co-authored-by: Emil Breiner <emil.breinr@krumedia.com>
The code added in the previous commit is only relevant for the mirror
format. With the long format, directories are created correctly in the
rotation routine. For the short format, the directory is initially
created, but only for a local destination, so adding a short snippet to
the rotation routine does it.
Before it was only possible to backup root directories (depth one, eg.
/etc, /var, /usr, …). For example consider the following truncated toy
configuration for rsync:
[general]
mountpoint = /media/
backupdir = backup
[source]
include = /foo/bar
include = /etc
We assume you have prepared the existent but empty base-directory
/media/backup/ at your destination, then against the expectation, the
following error rises:
rsync: mkdir "/media/backup/foo/bar/" failed: No such file or directory (2)
The reason is that rsync will take the files from the folder /foo/bar/
and tries to put them into the ulitmate – and therefore existent –
destination /media/backup/foo/bar. This is not what you would expect,
since the prepared directory /media/backup exists and the application
should create the rest itself.
This Merge Request adds a little script which creates the required
directories in the destination before running rsync. It assumes that
the base-directory already exists, so that not too much directories
are created automatically (could be a mess if you accidentally type the
wrong path)
