Book Image

PostgreSQL 11 Administration Cookbook

By : Simon Riggs, Gianni Ciolli, Sudheer Kumar Meesala
Book Image

PostgreSQL 11 Administration Cookbook

By: Simon Riggs, Gianni Ciolli, Sudheer Kumar Meesala

Overview of this book

PostgreSQL is a powerful, open source database management system with an enviable reputation for high performance and stability. With many new features in its arsenal, PostgreSQL 11 allows you to scale up your PostgreSQL infrastructure. This book takes a step-by-step, recipe-based approach to effective PostgreSQL administration. The book will introduce you to new features such as logical replication, native table partitioning, additional query parallelism, and much more to help you to understand and control, crash recovery and plan backups. You will learn how to tackle a variety of problems and pain points for any database administrator such as creating tables, managing views, improving performance, and securing your database. As you make steady progress, the book will draw attention to important topics such as monitoring roles, backup, and recovery of your PostgreSQL 11 database to help you understand roles and produce a summary of log files, ensuring high availability, concurrency, and replication. By the end of this book, you will have the necessary knowledge to manage your PostgreSQL 11 database efficiently.
Table of Contents (19 chapters)
Title Page
Copyright and Credits
About Packt

Using optimistic locking

If you perform work in one long transaction, the database will lock rows for long periods of time. Long lock times often result in application performance issues because of long lock waits:

SELECT * FROM accounts WHERE holder_name ='BOB' FOR UPDATE;
<do some calculations here>
UPDATE accounts SET balance = 42.00 WHERE holder_name ='BOB';

If that is happening, then you may gain some performance by moving from explicit locking (SELECT ... FOR UPDATE) to optimistic locking.

Optimistic locking assumes that others don't update the same record, and checks this at update time, instead of locking the record for the time it takes to process the information on the client side.

How to do it…

Rewrite your application so that the SQL is transformed into two separate transactions, with a double-check to ensure that the rows haven't changed (pay attention to the placeholders):

SELECT A.*, (A.*::text) AS old_acc_info
FROM accounts a WHERE holder_name ='BOB';