EDB Technical Update for Postgres Replication Server 7.0 (Limited Availability)

This article was originally published on June 03, 2019

• Single-Master and Multi-Master Replication Clusters: Configure a single master multi-replica cluster to offload reporting applications from production databases. Additionally, provision high-availability and write-scalability use cases with multiple master nodes that are synced in near-real time.
• Multiple Publications per Database: Create multiple publications in the same (SMR or MMR) cluster to create separate logical groups of tables.
• Hybrid replication: Setup a replication cluster such that each participating node is capable of acting as a master or replica or in both modes.
• High-Availability with Auto-Failover: Eliminate single-point of failure with multiple in-sync replicas managed by Apache Kafka and ZooKeeper.
• Offline Snapshot: The offline mode allows new replication nodes to quickly copy initial data for a large size database using PostgreSQL dump/restore utilities or other 3rd party options like Google Snapshot.
• Data Filtering: Define and apply row-level filters to control what data is replicated to target consumers.
• End-to-end Data Compression: Reduce network overhead and improve data transmission for large data sets through use of gzip/snappy data compression options.
• Command-line Utility: Use Rep-CLI (REST) client utility to automate cluster configuration.
• Basic Cluster-wide Monitoring: Health-check by monitoring key metrics and email alert notification based on metric threshold.
• Certificate-based Authentication: Configure SSL-based user authentication.
• ACL-based Access Control: Define users and roles for flexible and well-managed user access to the system. 
• Encrypted Database Connection: Secure data in transit for replication by enabling encrypted communication to the database server.
• Postgres v11, v10, v9.6 Support: Replication between PostgreSQL and EDB Postgres Advanced Server v11, v10, and v9.6 databases are supported.

• Management User Interface
• Column level filtering
• Update/Update and Insert/Insert conflict management
• Enhanced cluster-wide monitoring
• Windows Server support

• Replication fails with case-sensitive Publication table name
• In case of extended network failure, the EPRS server is subject to stop and needs to be restarted to join back the cluster
• The replicationlag command reports incorrect stats for a database that is stopped
• A non-admin user cannot execute createpub, joinpub and startstreaming RepCLI commands
• Replication fails for an ora-style partition table created in EPAS version 9.6
• Snapshot fails if Publication table filter rule contains modulus operator/absolute value, “@” or “%” character
• With the default configuration options, Snapshot might fail in a high-latency cross-region AWS cluster. The work-around is to increase the request.timeout.ms time in consumer.properties e.g. to 120000, and then perform Snapshot.