Added the mt-st package to the bacula-base Dockerfile to resolve issue with autochanger

[standekker]

While trying to get my Dell PowerVault 124T working with the Docker files in this repository, I noticed that I was not able to complete the autochanger test that is included in the btape tool, as recommended by the Bacula documentation

The btape tool autochanger command would output the following:

=== Autochanger test ===

3301 Issuing autochanger "loaded" command.
Nothing loaded in the drive. OK.

While there was a drive loaded in drive number 0. As Bacula was wrongfully assuming that drive 0 was empty, it would try to load a drive from slot 1 into drive 0 , which would return the error Drive 0 Full (Storage Element 1 loaded), while the btape tool would continue reporting

3303 Issuing autochanger "load 1 0" command.

Until a timeout was reached and the autochanger test failed.

The btape autochanger program calls a script /opt/bacula/scripts/mtx-changer in the bacula-sd container, which is a wrapper for the generic mtx command. This script is used to do all the interactions with the autochanger, acting as a proxy for mtx.

When the btape autochanger command is ran, it calls the mtx-changer script with the loaded argument to probe if the drive is loaded. It expects a return value of 0 if the drive is empty, and a return value of 1 if the drive is full. On my setup, it issued the following command:

./mtx-changer /dev/sg8 loaded 1 /dev/st0 0

Where dev/sg8 is the autoloader SCSI device and dev/st0 is the autoloader built-in SCSI tape drive.

When btape tries to run this command inside the bacula-sd container, the output is not a pure 1 or 0, but it also contains an error:

[root@86797ec3f40e scripts]# ./mtx-changer /dev/sg8 loaded 1 /dev/st0 0
/opt/bacula/scripts/mtx-changer.conf: line 84: mt: command not found
1

As the return value now contains an error besides the return value of 1, the btape program wrongfully interprets this, thinking the drive is empty. It then tries to load a tape into the full drive, which results in the autochanger test failing.

Note in the output above that the command is working correctly, as it returns a 1 while my drive was loaded at the time, it just also contains the error about the mt command, causing btape to interpret that the drive is empty
while it is actually full.

The /opt/bacula/scripts/mtx-changer.conf is a configuration file for the mtx-changer script and tries to figure out what version of mt is installed on line 84 and 89 of the mtx-changer.conf file, which is where the error is generated. As the mt-st package is not installed in the bacula-sd container, the error command not found is generated, which causes the issues down the line.

This pull request adds the mt-st repository to the bacula-base Dockerfile, which prevents the error from popping up when using the mtx-changer script, which fixes the issue where the autochanger test is not able to complete because of the command not found error being wrongfully interpreted by btape as the autochanger having an empty drive, to which it tries to load a drive into an already occupied drive, failing the test.

The fact that btape wrongfully interprets the error generated by the the mtx-changer script as the autochanger having an empty drive (which always will indicate an empty drive, regardless of the autochanger state if the mt-st package is not installed), is worth the bug report on the Bacula side, which I will try to file after this pull request.