Deadlocks

Some combinations of locks create a situation where some transactions block each other. Since all the involved transactions are waiting for a lock to be released, they will never unblock naturally. Such situations are called deadlocks. InnoDB (like other storage engines which need to deal with such situations) has an internal mechanism to detect deadlocks. It unblocks the situation by terminating the transaction that inserted, deleted, or updated the least number of rows. Such a transaction is called the victim.

Sometimes, a storage engine that uses table-level locks may be involved in a deadlock, which also involves InnoDB tables. In that case, InnoDB can only detect the deadlock if the innodb_table_locks server variable is ON.

In an environment where concurrency exists, deadlocks are normal. When a deadlock occurs, one or more transactions terminate with a 1213 error; this is not a concern, and applications should simply handle this error by restarting the transaction. Of course...