The DB_ENV->failchk() method now has a new option to notify environment processes of an application failure as soon as possible. This option causes any thread waiting on a mutex owned by a crashed process to return DB_RUNRECOVERY instead of hanging.
This option is enabled at compile time with the new configure option --enable-failchk_broadcast. When enabled, this option provides a new timeout. It also provides new events that notify the application when a thread is about to return DB_RUNRECOVERY.
timeout
parameter value for the
DB_ENV->set_timeout() method to indicate how long a thread
will wait for a mutex lock before checking whether
DB_ENV->failchk() has marked the mutex as failed.