Monitoring and logging setup for external databases

External PostgreSQL database systems have different logging options for monitoring performance and troubleshooting, however they are not enabled by default. In this section we provide the recommendations for self-managed PostgreSQL, and recommendations for some major providers of PostgreSQL managed services.

Recommended PostgreSQL Logging settings

You should enable the following logging settings:

• log_statement=ddl: log changes of database model definition (DDL), such as CREATE, ALTER or DROP of objects. This helps track recent model changes that could be causing performance issues and identify security breaches and human errors.
• log_lock_waits=on: log of processes holding locks for long periods, a common cause of poor query performance.
• log_temp_files=0: log usage of intense and unusual temporary files that can indicate poor query performance.
• log_autovacuum_min_duration=0: log all autovacuum executions. Autovacuum is a key component for overall PostgreSQL engine performance. Essential for troubleshooting and tuning if dead tuples are not being removed from tables.
• log_min_duration_statement=1000: log slow queries (slower than 1 second).

The full description of these parameter settings can be found in PostgreSQL error reporting and logging documentation.

Amazon RDS

The Amazon Relational Database Service (RDS) provides a large number of monitoring metrics and logging interfaces. Here are a few you should configure:

• Change all recommended PostgreSQL Logging settings through RDS Parameter Groups.
  • As the recommended logging parameters are dynamic in RDS you don't require a reboot after changing these settings.
  • The PostgreSQL logs can be observed through the RDS console.
• Enable RDS performance insight allows you to visualise your database load with many important performance metrics of a PostgreSQL database engine.
• Enable RDS Enhanced Monitoring to monitor the operating system metrics. These metrics can indicate bottlenecks in your underlying hardware and OS that are impacting your database performance.
  • In production environments set the monitoring interval to 10 seconds (or less) to capture micro bursts of resource usage that can be the cause of many performance issues. Set Granularity=10 in the console or monitoring-interval=10 in the CLI.