Stratégies de reprise après sinistre

Even with high availability (HA) setups like replication, unexpected disasters can strike. Think data center outages, major corruption, or human error.

Disaster Recovery (DR) is your plan B. It's about recovering your database system after a catastrophic event to minimize data loss and downtime.

It's easy to confuse High Availability (HA) with Disaster Recovery (DR), but they serve different purposes:

• HA: Aims to keep systems running with minimal interruption during minor failures (e.g., a single server crash, network glitch). It often involves automatic failover to a standby.
• DR: Focuses on recovering from major, widespread failures that take down an entire region or multiple components, where HA alone might not suffice.

Replication is a cornerstone for both, but DR requires additional planning for recovery.

A robust Disaster Recovery plan for PostgreSQL involves several key components:

• Regular Backups: Both full base backups and continuous Write-Ahead Log (WAL) archiving.
• Recovery Objectives: Defining your RTO (Recovery Time Objective) and RPO (Recovery Point Objective).
• Offsite Storage: Storing backups safely away from your primary data center.
• Tested Procedures: Documented and regularly practiced restore processes.

These two metrics are crucial for any DR plan:

• Recovery Time Objective (RTO): This is the maximum acceptable duration of time that a computer system, application, or network can be down after a disaster. It's about time to recover.
• Recovery Point Objective (RPO): This is the maximum acceptable amount of data loss measured in time. For example, an RPO of 1 hour means you can lose up to 1 hour of data. It's about data loss tolerance.

Your RTO and RPO will dictate your backup frequency and recovery strategy.

For replicated environments, you typically combine two types of backups:

• Base Backup: A full snapshot of your database cluster at a specific point in time. This forms the foundation for any restoration.
• WAL Archiving: Continuously saving the Write-Ahead Log (WAL) files. These logs record all changes to your database and are essential for point-in-time recovery and keeping standbys up-to-date.

Together, they enable restoring to any point within your archived history.

pg_basebackup is the go-to tool for creating a base backup. It can even take a backup from a running standby server to offload work from the primary!

This command creates a full copy of the data directory. The -X stream or -X fetch options are often used to include WAL files, making the backup self-contained.

Try this example:

WAL archiving is critical for point-in-time recovery and ensuring your standbys can catch up even if the primary crashes. It involves copying completed WAL files to a safe, often remote, location.

You configure this in your postgresql.conf file:

If your primary server fails catastrophically, you'll need to restore it. This typically involves:

1. Initializing a new data directory.
2. Restoring the latest base backup into the new directory.
3. Configuring a recovery.signal file (or standby.signal for older versions) and postgresql.conf to point to your WAL archive.
4. Starting PostgreSQL, which will then replay the WAL files to bring the database to the desired point in time.

This process can also be used to create a new primary from a standby backup.

After a disaster, you might also need to rebuild your standby servers to connect to the newly restored (or promoted) primary. The process is similar to setting up a new standby:

1. Take a fresh base backup from the new primary.
2. Configure the standby's postgresql.conf and standby.signal to connect to the new primary.
3. Start the standby. It will then stream WAL from the new primary to synchronize.

This ensures your replication chain is healthy again.

A DR plan is only as good as its last test. Regularly practicing your recovery procedures is vital. This helps you:

• Identify gaps or errors in your documentation.
• Ensure your recovery time (RTO) is achievable.
• Train your team on the recovery process.
• Confirm backups are valid and restorable.

Schedule regular DR drills, perhaps annually or semi-annually, in a non-production environment.

Imagine a major data center outage has taken down your primary PostgreSQL server and all its local standbys. You have offsite base backups and continuous WAL archiving.

Which of the following describes the maximum acceptable amount of data loss you are willing to tolerate in this scenario?

We've covered the essentials of Disaster Recovery for PostgreSQL in replicated environments. Remember these key takeaways:

• DR protects against catastrophic failures, complementing High Availability.
• Define RTO (time to recover) and RPO (data loss tolerance) for your systems.
• Combine base backups and WAL archiving for comprehensive recovery.
• pg_basebackup and archive_command are crucial tools.
• Always test your DR plan regularly to ensure readiness.

A well-practiced DR plan is your safety net for critical data.