Glossary · System coordination, integration and orchestration
Replication lag
Also known as: Replica lag
German: Replikationsverzögerung
In distributed systems and databases, replication lag is the delay between a change being made on a primary node and the same change becoming visible on a replica; during this time, readers of the replica see outdated data.
- System integration
In one sentence
Replication lag is the delay until a change on a primary node becomes visible on a replica, during which the replica shows outdated data.
Example
A dashboard reading from a historian replica shows a batch as still running for several seconds after the primary has recorded its completion.
How it applies
- Engineering: Asynchronous replication gives better performance and availability but causes lag; synchronous replication avoids lag at the cost of latency. Decide per use case which data may be read from replicas.
- Operation: Monitor lag and alert when it exceeds a threshold. High lag at the moment of a Switchover means recent changes can be lost.
- Integration: Applications that write and immediately read back must read from the primary or use read-your-writes guarantees, or users will see their own changes disappear.
- Documentation: State in the system documentation which data is replicated asynchronously, the expected lag and the possible data loss window on failover (recovery point objective).
Replication lag vs. network latency
Network latency is the transport time of a message. Replication lag also includes queuing and processing on the replica and can grow to minutes under load, even on a fast network.