Crash Recovery

MySQL Server supports crash recovery, which ensures durability and enables data recovery in the event of an unexpected server exit.

While this redundancy is advantageous during normal operation of the server, it can lower the performance of large data imports. You can disable the crash recovery processes, temporarily, enabling you to execute DML statements without the overhead of synchronization.
Note

If any component of a standalone DB system fails while crash recovery is disabled, the DB system enters a FAILED state and is unrecoverable. It is recommended to perform a full manual backup before disabling crash recovery. Highly available DB systems in multi-availability domains are more failure resistant but in certain circumstances, can also become unrecoverable.

Disabling crash recovery disables the following:

  • InnoDB redo log
  • Doublewrite buffer
  • Binary log synchronization

If you disable crash recovery, you cannot use the following HeatWave Service processes:

  • Backups (manual and automatic)
  • DB system stop and restart
Note

It is not recommended to run a DB system without crash recovery except when performing large imports of data.

If crash recovery is disabled when a DB system upgrade operation begins, it is re-enabled for the duration of the upgrade process, and disabled again when the upgrade is finished. The same is true during a failover of a highly available DB system primary to secondary. When the promotion process is complete, crash recovery is disabled again.