Take advantage of PostgreSQL 18 engine capabilities with the Citus 14-powered elastic cluster horizontal scale-out experience.
PostgreSQL 18 meets managed horizontal scale
We’re proud to preview PostgreSQL 18 for elastic clusters in Azure Database for PostgreSQL Flexible Server. Powered by Citus 14, elastic clusters bring PostgreSQL 18 into the managed elastic-cluster experience, allowing teams to evaluate the latest PostgreSQL engine together with distributed SQL, sharding, and horizontal scale-out. The result is a more predictable path for PostgreSQL 18 applications leveraging sharded data, including applications that rely on newer SQL semantics.
This preview is an opportunity to test application compatibility, validate distributed query behavior, assess operational workflows, and share feedback before making production decisions. As with any preview, customers should review current documentation for supported features and limitations rather than assume that every capability in an open-source Citus 14.x release is exposed in the Azure service.
In addition to the performance improvements and new features included in the Postgres 18 release, elastic clusters for PG18 include added new benefits into distributed execution. We now preserve virtual and stored generated-column definitions and metadata, propagate temporal constraints, and support OLD and NEW values in distributed RETURNING. We also support PostgreSQL 18 JSON_TABLE COLUMNS in distributed queries, helping developers turn JSON documents into relational rows while continuing to work within a distributed data model.
Under the Hood
Supporting a new PostgreSQL major version in a distributed database involves more than compiling against the new engine. Your elastic cluster must correctly parse and deparse new syntax, propagate supported DDL and metadata, plan queries across shards, execute work on coordinator and worker nodes, and preserve PostgreSQL semantics throughout. Elastic clusters deliver end-to-end PostgreSQL 18 support across coordinator and workers while retaining the distributed SQL and sharding capabilities customers depend upon.
Observability also matters when a query spans nodes. Elastic clusters now carry PostgreSQL 18 EXPLAIN changes, including WAL fields, through distributed EXPLAIN output so teams can reason about both engine activity and distributed execution. We’ve also improved the citus_create_restore_point() for consistent restore points in MX or multi-writer clusters. In the Azure preview, DDL and cluster-wide operations are performed through the coordinator and teams should validate their deployment and operational procedures against the documented managed-service behavior.
Workloads to evaluate
The preview can be useful anywhere teams need to understand how a PostgreSQL 18 application behaves when data and queries are distributed. Four workload patterns are especially relevant:
- Multi-tenant SaaS platforms that distribute by tenant identifier to keep related rows co-located and scale capacity as the tenant base grows.
- Real-time analytics that aggregate recent or operational data and can benefit when eligible queries run in parallel across shards.
- High-ingest event platforms that can test uuidv7() locality, PostgreSQL 18 I/O improvements, time-oriented indexes, and distributed write behavior together.
- Microservices and business applications using schema-based sharding to preserve familiar per-tenant schemas while managing the cluster as one Azure resource.
Use the preview to answer production questions
A useful preview evaluation starts with representative schemas, data sizes, and query mixes. Confirm that extensions, drivers, ORMs, and migration tooling are compatible with PostgreSQL 18. Review shard-key choices and colocation, then compare single-shard, cross-shard, and parallel query plans. Exercise node addition and online data rebalancing and observe application behavior while data placement changes.
Test database semantics as carefully as performance. If an application adopts virtual generated columns, temporal constraints, JSON_TABLE, or OLD and NEW values in RETURNING, verify those paths on distributed tables as well as local ones. Use the richer EXPLAIN output to establish a baseline and identify whether time is spent in the PostgreSQL engine, distributed execution, data movement, or application round trips.
Finally, evaluate the managed experience on its documented terms. The preview makes PostgreSQL 18 available through elastic clusters, while Azure controls which service features and operations are exposed. Preview findings and customer feedback will help shape future readiness and ensure your applications are able to function at peak performance when the general availability data is announced.
Get started and learn more
Create an elastic cluster and begin a focused compatibility and workload evaluation. The following public resources provide current service guidance and upstream technical details:
Create an elastic cluster: https://learn.microsoft.com/en-us/azure/postgresql/elastic-clusters/create-elastic-cluster
Elastic clusters concepts: https://learn.microsoft.com/en-us/azure/postgresql/elastic-clusters/concepts-elastic-clusters
PostgreSQL 18 release announcement and release notes: https://www.postgresql.org/docs/current/release-18.html
Citus 14 release overview: https://www.citusdata.com/updates/v14-0/
We encourage PostgreSQL developers, architects, SaaS builders, and data platform teams to test the preview with representative workloads and share feedback on compatibility, distributed behavior, and the managed experience.
Special thanks to the entire Elastic Clusters team for their incredible effort in this release.