Release notes for EDB Postgres Distributed version 5.0.0 v5

Redefined Always-ON to support wider variety of deployments.

BDR cluster elects a write leader for every group (and associated location) using per group Raft when routing is enabled for the group. It takes care of write leader failover and provides SQL commands to change a write leader.

EDB Advanced Storage Pack provides advanced storage options for PostgreSQL databases in the form of table access method (TAM) extensions. These storage options can enhance the performance and reliability of databases without requiring application changes.

The durability options such as Group Commit, CAMO, Eager Replication or Lag Control are now all configured through Commit Scope configuration.

EDB Postgres Distributed 5 fully supports the Transparent Data Encryption feature in EDB Postgres Advanced and EDB Postgres Extended

BDR extension can now send monitoring metrics as well as traces to the OpenTelemetry collector for better integration with existing monitoring solutions.

EDB Postgres Distributed 5 is compatible with Postgres 12 to 15.

The bdr.worker_errors and bdr.state_journal_details view were replaced by unified bdr.event_summary which also include changes in Raft role for the local node. In the future additional events may be added to it.

Transaction tracking now uses shared memory instead of bdr.internal_node_pre_commit catalog which considerably improves performance as it does not incur additional I/O.

Allows Decoding Worker feature to be used in clusters using non-default replication sets like asymmetric replication setup.

Extend autopartition/autoscale to support HASH partitioning. Many of things that are required for RANGE partitioning are not needed for HASH partitioning. For example, we expect to create all HASH partitions in one go (at least for the current work; later we may change this). We don't expect HASH partitions to be moved to a different tablespace or dropped. So data retention policies don't apply for HASH partitioning.

The utility supports benchmarking CAMO transactions and in future releases will be used for benchmarking PGD specific workloads.

In this release, the autopartition work queue mechanism has been moved to a separate module called Task Manager (taskmgr). The task manager is responsible for creating new tasks and executing the ones created by the local node or the task manager leader node. The autopartition worker is thus renamed as taskmgr worker process in this release.

In the older PGD releases, the Raft leader was responsible for creating new work items. But that creates a problem because a witness node can become a Raft leader while it does not have the full view of the cluster objects. In this release, we have introduced a concept of Task Manager Leader node. The node is selected automatically by PGD, but for upgraded clusters, its important to set the node_kind properly for all nodes in the cluster. The user is expected to this manually after upgrading to the latest PGD version by calling bdr.alter_node_kind() SQL function for each node.

This better differentiates different kinds of nodes such as data, witness, subscriber-only and standby.

A TCP layer 4, pass through proxy for PGD cluster using routing capabilities of BDR.

CLI supports two new commands verify-settings and verify-cluster. verify-settings verifies the PostgreSQL configuration of each node in a PGD cluster against the recommendations. verify-cluster verifies the PGD cluster architectures against the flexible architecture deployment recommendations.

pgd supports create-proxy, delete proxy, set-group-options, set-node-options, set-proxy-options, show-proxies, show-groups and switchover to configure and manage Proxy per group.