Am 07.06.2017 um 12:34 schrieb Timothy D. Legg:
So I stop the mysql process and then for the sake of recoverability, I tar up /var/lib/mysql/mywiki into a file called mywiki_without_new_data.tar. I then rm -rf /var/lib/mysql/mywiki and then untar the mywiki_with_new_data.tar in it's place. MariaDB will no longer start. If I again rm -rf /var/lib/mysql/mywiki and then untar mywiki_without_new_data.tar into the location, MariaDB starts without error.
This isn't a code blue situation, my friend can recreate the change he made and that is the easiest solution, but others reading this might not have that luxury. I am also quite curious about what went wrong and how to repair the problem. I didn't think this was possible to fail.
There is an interesting except from the errors described below. I'm not certain what exactly this means, or how to correct it. Given that MariaDB could have failed uncleanly when the mysql database was imported, recovery may be infeasible. (the wiki connected to MariaDB as root, it's bad, I know)
<excerpt /var/lib/mysql/altdorf.err> InnoDB: Error: tablespace id is 235 in the data dictionary InnoDB: but in file ./mywiki/objectcache.ibd it is 254! 2017-06-07 11:45:08 5c7ff420 InnoDB: Assertion failure in thread 1551889440 in file fil0fil.cc line 638 InnoDB: We intentionally generate a memory trap. </excerpt>
you can't do that with InnoDB databases beaus eby design there exists a idiotic global table space even with "innodb_file_per_table" - why the whole server needs to crash with 10.1 and newer in such cases is a different story