Managing geo-replication

PGD supports clusters that span multiple geographic locations, including data centers, availability zones, and cloud regions, where each location can accept writes independently. With geo-replication, you can build resilient, globally distributed database systems that protect against location failures and keep data close to users.

When to use geo-replication

Geo-replication is the right choice when you need one or more of the following:

• Location failure protection: if a data center or availability zone goes offline, another location continues to serve reads and writes without data restore.

• Data residency: regulatory requirements mandate that certain data stays within a specific region or country.

• Low-latency reads across regions: users in different regions need fast access to data without round-tripping to a central location.

• Disaster recovery with low Recovery Time Objective (RTO): you need near-instantaneous failover to another location without manual restore from backup.

Deployment workflow

Follow these steps when setting up a new geo-distributed PGD cluster.

1. Choose a deployment pattern

Decide on the number of active write locations, routing mode, and commit scope before configuring anything. The deployment pattern affects every subsequent decision.

See Choosing a geo-replication deployment pattern .

2. Plan your node distribution and create subgroups

Before provisioning nodes, measure network latency between locations and plan quorum so that no single location holds a majority of voting nodes. Determine whether you need a witness location. Then create a subgroup for each location under the top-level group and assign nodes to their subgroups.

See Setting up your geo-replication topology .

3. Configure routing

Decide between local routing (write leader per location) and global routing (single cluster-wide write leader). Local routing is the default and requires no changes. For global routing, enable it on the top-level group and disable it on subgroups.

Set route priority to control which nodes are preferred as write leaders during failover.

See Configuring routing thresholds .

4. Configure commit scopes

Choose a commit scope based on your durability requirements and configure it for each location subgroup.

See Migration to commit scopes .

5. Tune replication and sequences

Tune parallel apply and streaming settings to reduce replication lag under high write volumes. Confirm that sequences are configured to generate cluster-wide unique values.

See Replication and sequence tuning .

6. Configure lag control

If cross-region replication lag is a concern, add Lag Control rules to your commit scope to throttle write throughput when remote nodes fall behind.

See Controlling replication lag .

7. Verify the cluster and prepare for operations

Once configuration is complete, verify that subgroups, routing, and commit scopes are set up correctly before going to production. The maintenance page also covers ongoing operational tasks such as rolling upgrades, DDL considerations, network partition handling, and conflict monitoring.

See Performing cluster maintenance .

• Choosing a geo-replication deployment pattern
  • Deciding on the number of active write locations
  • Choosing between local and global routing
  • Enforcing data residency requirements
  • Comparison
• Setting up your geo-replication topology
  • Planning your node distribution
  • Creating location subgroups
  • Assigning nodes to subgroups
• Configuring routing
  • Configuration examples
    • Local routing
    • Global routing
  • Tuning Connection Manager
  • Setting route priority
  • Verifying current settings
• Configuring commit scopes for geo-replication
  • Choosing a commit scope
  • Creating the commit scope
    • Creating a Quorum Commit scope
    • Creating a CAMO scope
  • Setting the default commit scope
  • Verifying the configuration
  • Avoiding common pitfalls
• Controlling replication lag
  • Using lag control
    • Configuring Lag Control
    • When Lag Control is not enough
  • Configuring routing thresholds
  • Tuning replication apply
  • Monitoring replication lag
  • Subscriber-only nodes
    • Adding a subscriber-only node
    • Routing reads to subscriber-only nodes
    • Limitations of subscriber-only nodes
• Replication and sequence tuning
  • Parallel Apply
  • Apply delay
  • Sequences
• Performing cluster maintenance
  • Verifying the cluster configuration
  • Managing database changes
    • Running DDL
    • Running VACUUM and ANALYZE
  • Managing write leadership
    • Handling failover
    • Maintaining quorum
    • Performing a planned switchover
  • Performing node maintenance
    • Rolling operations across locations
    • Fencing nodes
  • Monitoring the cluster and responding to incidents
    • Handling network partitions
    • Monitoring conflicts
    • Configuring conflict resolvers