Book Image

Learn PostgreSQL - Second Edition

By : Luca Ferrari, Enrico Pirozzi
1 (2)
Book Image

Learn PostgreSQL - Second Edition

1 (2)
By: Luca Ferrari, Enrico Pirozzi

Overview of this book

The latest edition of this PostgreSQL book will help you to start using PostgreSQL from absolute scratch, helping you to quickly understand the internal workings of the database. With a structured approach and practical examples, go on a journey that covers the basics, from SQL statements and how to run server-side programs, to configuring, managing, securing, and optimizing database performance. This new edition will not only help you get to grips with all the recent changes within the PostgreSQL ecosystem but will also dig deeper into concepts like partitioning and replication with a fresh set of examples. The book is also equipped with Docker images for each chapter which makes the learning experience faster and easier. Starting with the absolute basics of databases, the book sails through to advanced concepts like window functions, logging, auditing, extending the database, configuration, partitioning, and replication. It will also help you seamlessly migrate your existing database system to PostgreSQL and contains a dedicated chapter on disaster recovery. Each chapter ends with practice questions to test your learning at regular intervals. By the end of this book, you will be able to install, configure, manage, and develop applications against a PostgreSQL database.
Table of Contents (22 chapters)
20
Other Books You May Enjoy
21
Index

Deadlocks

A deadlock is an event that happens when different transactions depend on each other in a circular way. Deadlocks are, to some extent, normal events in a concurrent database environment and nothing an administrator should worry about, unless they become extremely frequent, meaning there is some dependency error in the applications and the transactions.

When a deadlock happens, there is no choice but to terminate the locked transactions. PostgreSQL has a very powerful deadlock detection engine that does exactly this job: it finds stalled transactions and, in the case of a deadlock, terminates them (producing a ROLLBACK).

In order to produce a deadlock, imagine two concurrent transactions applying changes to the very same tuples in a conflicting way. For example, the first transaction could do something like the following:

-- session 1
forumdb=> BEGIN;
BEGIN
forumdb=> SELECT txid_current();
 txid_current
--------------
         4875
(1 row)
forumdb=> UPDATE...