Data are no longer lost when the *MariaDB* init script fails
Previously, if the *MariaDB* init script failed, it called "rm -rf" on the whole directory. This could consequently lead to a loss of data or even deletion of the mount point. With this update, several additional checking mechanisms have been added to the init script. Now, if the script fails, it removes only files newer than a timestamp generated prior to the critical file operations. In addition, a set of human-readable status reports and error messages have been added.
Short sum-up, consider that:
* /var/lib/mysql is mount point
* 'systemctl start mariadb' is executed
* this runs '/usr/bin/mysql_install_db' in /var/lib/mysql when
the '[ ! -d /var/lib/mysql/mysql ]'
* if the mysql_install_db fails, 'rm -rf' is called there
The issue is when user mistakenly mounts the directory wrongly.
E.g. the whole '/var' mountpoint mounted as /var/lib/mysql:
* the '/var/lib/mysql/mysql' does not exist
* the mysql_install_db is called, but it fails (permissions)
* the 'rm -rf' is called, removes recursivelly all files owned by mysql
Clear user error, but we can protect user against disaster.
For more info:
+++ This bug was initially created as a clone of Bug #1335849 +++
There are other cases too:
- backup of databases, for example
mkdir /var/lib/mysql/BACKUP; mv /var/lib/mysql/* /var/lib/mysql/BACKUP
- permission problems or fail of filesystem + over quota = removal of files
- may be others too...
Should be fixed before more users will loss its data.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHSA-2017:2192