Memahami Kunci dan Deadlock

When many users or applications access a database at the same time, it's called concurrency. PostgreSQL, like other databases, needs to manage these concurrent operations carefully.

Without proper management, multiple operations could try to modify the same data simultaneously, leading to inconsistencies or errors. This is where locks come in!

A database lock is a mechanism that controls access to data by multiple transactions. It ensures that data remains consistent and prevents conflicts.

• Data Integrity: Guarantees that data is accurate and reliable.
• Consistency: Ensures that transactions see a consistent view of the database.
• Conflict Prevention: Stops different operations from corrupting data by trying to write to the same place at the same time.

PostgreSQL uses various lock modes to define the level of access transactions have to a particular resource. These modes determine compatibility: some allow shared access, while others are exclusive.

• ACCESS SHARE: Acquired by simple SELECT statements. Allows other transactions to also acquire ACCESS SHARE, ROW SHARE, etc.
• ROW EXCLUSIVE: Acquired by INSERT, UPDATE, DELETE. Allows concurrent reads but prevents other transactions from acquiring ROW EXCLUSIVE on the same row.
• ACCESS EXCLUSIVE: The most restrictive lock. Acquired by DROP TABLE or TRUNCATE. Prevents all other access to the table.

Most of the time, PostgreSQL automatically acquires the necessary locks for you. These are called implicit locks.

For example:

• When you run a SELECT query, an ACCESS SHARE lock is acquired on the table.
• When you run an UPDATE, INSERT, or DELETE query, a ROW EXCLUSIVE lock is acquired on the affected rows and a corresponding table-level lock.

This automatic locking ensures data consistency without you needing to explicitly manage it.

While implicit locks handle most cases, you can also acquire locks explicitly using the LOCK TABLE command. This is useful for specific scenarios where you need to control access more precisely.

You can specify the lock mode, like ACCESS EXCLUSIVE to block all other operations, or SHARE for concurrent index creation.

Try running this example. It creates a simple table and then explicitly locks it in ACCESS EXCLUSIVE mode. While this lock is held (for the duration of the transaction), no other transaction can read or write to this table.

For fine-grained control, you can lock specific rows using SELECT FOR UPDATE or SELECT FOR SHARE. This is crucial for preventing race conditions in application logic.

• SELECT ... FOR UPDATE: Acquires a ROW EXCLUSIVE lock on selected rows. Other transactions can read but cannot update or lock these rows for update until your transaction commits.
• SELECT ... FOR SHARE: Acquires a ROW SHARE lock. Other transactions can also acquire ROW SHARE locks, but not ROW EXCLUSIVE.

To see what locks are currently active in your database, you can query the pg_locks system view. This view provides detailed information about each lock.

Key columns to look for:

• pid: The process ID holding or waiting for the lock.
• locktype: Type of resource being locked (e.g., relation, transactionid).
• mode: The lock mode (e.g., ACCESS SHARE, ROW EXCLUSIVE).
• granted: t if the lock is held, f if waiting.

You can query it like this:

A deadlock occurs when two or more transactions are waiting for each other to release a resource, resulting in a standstill. Neither transaction can proceed.

PostgreSQL automatically detects deadlocks and aborts one of the transactions (the 'deadlock victim') to allow the other to complete. This usually results in an error message for the aborted transaction.

Imagine two transactions, T1 and T2, trying to update two rows, Row A and Row B:

• T1: Locks Row A, then tries to lock Row B.
• T2: Locks Row B, then tries to lock Row A.

At this point, T1 holds A and waits for B, while T2 holds B and waits for A. Neither can progress. PostgreSQL detects this cycle and terminates one transaction.

We've covered the essentials of database locks and deadlocks in PostgreSQL. Locks ensure data consistency and integrity in concurrent environments, operating through various lock modes.

Remember:

• PostgreSQL uses implicit and explicit locks.
• pg_locks helps monitor active locks.
• Deadlocks occur when transactions wait for each other, and PostgreSQL detects and resolves them by aborting one transaction.

Understanding these concepts is key to building robust and performant applications!