Resolve InnoDB dictionary or tablespace metadata corruption

Issue

The database crashes or reboots unexpectedly when running DDL operations, such as ALTER TABLE ... DROP FOREIGN KEY.

The error log contains assertion failures similar to:

InnoDB: Assertion failure in file .../btr0btr.cc line 204
InnoDB: Failing assertion: mach_read_from_4(seg_header + FSEG_HDR_SPACE) == space

Environment

MariaDB Enterprise Server

Cause

The InnoDB data dictionary or tablespace metadata has become corrupted.

Resolution

To resolve the corruption, rebuild the InnoDB system tablespace by exporting the data, removing the corrupted system files, and restoring the data.

⚠️ Warning: This is a destructive process. Ensure you have a complete, verified backup of all databases before proceeding.

1. Export all data, routines, and events using mariadb-dump:

   mariadb-dump --all-databases --routines --events > /tmp/all_dbs.sql

2. Drop the affected databases. This step prevents conflicts and errors (such as "Table doesn't exist" or "Directory not empty") during the restore process.
3. Stop the MariaDB service:

   systemctl stop mariadb

4. Navigate to your MariaDB data directory and rename or remove the InnoDB system files:
   • ibdata1
   • ib_logfile0
   • ib_logfile1
   • ibtmp1 (if it exists)

   Example:

   mv ibdata1 ibdata1.bak
   mv ib_logfile0 ib_logfile0.bak
   mv ib_logfile1 ib_logfile1.bak
   mv ibtmp1 ibtmp1.bak

5. Start the MariaDB service. The server will initialize a clean, uncorrupted data dictionary and fresh system tablespaces:

   systemctl start mariadb

6. Restore the data from the dump file created in step 1:

   mariadb < /tmp/all_dbs.sql