MariaDB won’t start after php-updater change – “Unknown/unsupported storage engine: InnoDB”

Question

I am running Nextcloud on Debian 11. Recently, I tried upgrading using php-updater, and afterwards MariaDB stopped working properly.

System details:

Debian 11

MariaDB 11.3.2

Apache with mpm_prefork (no php-fpm)

Nextcloud data in /var/www/owncloud/ (I am in the middle of a migration. Owncloud to Nextcloud worked so far, however, I wanted to use the php-updater script for the following Nextcloud upgrades)

Database: owncloud (visible via SHOW DATABASES; when MariaDB works)

What happened:

Right before this issue happened, I used php-updater. It said

It looks like both the 'Apache 2.0 Handler SAPI' libapache2-mod-php7.4 and the 'FastCGI Process Manager SAPI' php7.4-fpm are connected to it. This is a serious misconfiguration which must be fixed.
I: The Apache2 server has loaded the 'prefork' multi-processing module required by the Apache 2.0 Handler SAPI
   and the 'event' module, which is required for operation with php-fpm, is not loaded.

I used the option disconnect apache2 from php*-fpm, without changing the loaded modules. , MariaDB takes minutes to stop/start, or fails to start.

Log shows:

[ERROR] Unknown/unsupported storage engine: InnoDB
[ERROR] Aborting

• My /var/lib/mysql/ currently contains:

ibdata1.bak (618M)
ibtmp1
undo001, undo002, undo003
(no ib_logfile0 / ib_logfile1 present anymore)

• I tried adding a low-memory config (60-lowmem.cnf):

[mysqld]
innodb_force_recovery=6
innodb_buffer_pool_size = 16M
innodb_log_file_size    = 16M
innodb_log_buffer_size  = 8M
max_connections         = 10
table_open_cache        = 100
tmp_table_size          = 8M
max_heap_table_size     = 8M
skip-name-resolve

• With innodb_force_recovery=6, MariaDB briefly started once, but after restoring configs, it fails again.

What I tried so far

• Restored ibdata1 from ibdata1.bak
• Tried creating fresh empty ib_logfile0 / ib_logfile1 (owned by mysql:mysql)
• Adjusted permissions for all /var/lib/mysql/*
• Tried different innodb_force_recovery levels (1–6)

Current state

• MariaDB always fails with Unknown/unsupported storage engine: InnoDB.
• I cannot get it to start reliably, even in recovery mode.
• Nextcloud is down because the database is inaccessible.

My questions

1. How can I force MariaDB to accept the existing ibdata1.bak (InnoDB system tablespace)?
2. Is there a safe way to recreate the missing ib_logfile* files so MariaDB will start?
3. If recovery is impossible, what’s the recommended way to salvage data from ibdata1.bak and the /var/lib/mysql/owncloud/ directory? I also have some backup data - at least before the Nextcloud migration, the relevant directories and a database backup were done. After the migration I backed up the data directory and once the database, when I once was able to restart the maria database.

Any hints or experience with recovering MariaDB after missing/corrupted InnoDB logs would be very much appreciated.

Answer 1

I was not able to solve the task by forcing MariaDB to accept the existing ibdata1.bak or recreate the ib_logfile*.

Solution that worked for me

In the process I restored from a clean backup. However, due to my new configuration or because it was wrongly configured elsewhere, I got a character set error after that. A workaround or solution of this issue helped.

0. Remove database datadir

   As root, I (re-)moved /var/lib/mysql/mysql.

   mv /var/lib/mysql/mysql /var/lib/mysql/mysql.broken.$(date +%s)
   

   Then

   mariadb-install-db --user=mysql --basedir=/usr --datadir=/var/lib/mysql
   
   systemctl start mariadb
   

1. Restore from a clean backup
   I restored from my latest mysqldump backup. I still have the old owncloud database user. Maybe the character set should have been configured here.

   mariadb -u root -p -e "DROP DATABASE IF EXISTS owncloud; CREATE DATABASE owncloud CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci;"
   mariadb -u owncloud -p owncloud < owncloud_backup.sql
   

   However, then I got

   # sudo -u www-data php occ maintenance:mode --on 
   An unhandled exception has been thrown: 
   Doctrine\DBAL\Exception: Failed to connect to the database: An exception occurred in the driver: SQLSTATE[HY000] [2054] Server sent charset unknown to the client. Please, report to the developers in /var/www/owncloud/lib/private/DB/Connection.php:139
   

   Before a new import, I replaced all utf8mb3 collations in the dump with plain utf8 (or utf8mb4):

   sed -i 's/utf8mb3/utf8/g' owncloud_backup.sql
   

2. Fix MariaDB server character set config
   In /etc/mysql/mariadb.conf.d/50-server.cnf, I reconfigured the character set:

   character-set-server = utf8mb4
   collation-server     = utf8mb4_general_ci
   skip-character-set-client-handshake
   

   Then restarted:

   systemctl restart mariadb
   

3. Verify Nextcloud connectivity
   After the config change, Nextcloud was able to connect again. Maintenance mode could be enabled/disabled:

   sudo -u www-data php occ maintenance:mode --on
   sudo -u www-data php occ maintenance:mode --off
   

   Login worked afterwards.

Key takeaways

• If MariaDB reports unknown/unsupported storage engine InnoDB, chances are the InnoDB system tablespace is corrupted beyond repair. Restore from mysqldump.
• With SQLSTATE[HY000] [2054] Server sent charset unknown to the client, check for utf8mb3 leftovers in your dump. One can try character-set-server=utf8mb4 is set in MariaDB config.
• Always keep a working backup routine — that’s what saved my Nextcloud instance here.

Hope this helps anyone hitting the same wall after an InnoDB crash.