Database Technology

AWS details building a high-concurrency async Python application using Tortoise ORM and Amazon Aurora DSQL. Key technical adaptations include implementing UUID primary keys and configuring IAM-authenticated asyncpg connections with a specific connection-pool patch. The guide also covers individual DDL execution and integrating optimistic concurrency control retry logic for robustness.

• Use UUIDs for primary keys to support distributed async workloads effectively.
• Patch the connection pool to enable IAM authentication with asyncpg.
• Execute DDL statements individually to ensure compatibility and control.
• Implement optimistic concurrency control retry logic to handle contention.

Percona Operator for PostgreSQL 3.1.0 introduces transparent data encryption, logical replicas for read scaling, and persistent logging within the custom resource. These features address critical compliance and performance requirements by enabling data-at-rest protection, offloading read traffic from the primary, and ensuring log durability. The release consolidates these capabilities directly into the operator configuration to streamline platform governance.

• Transparent Data Encryption (TDE) is now supported for compliance with data-at-rest security standards.
• Logical replicas allow read scaling without adding load to the primary database node.
• Persistent logging ensures log availability for auditing and troubleshooting across pod restarts.
• All features are configurable via the custom resource, reducing manual infrastructure overhead.

PostgreSQL 18 adds the max_active_replication_origins GUC, resolving a decade-long debt from a comment in version 9.5. Previously, the limit on tracking replication origins was borrowed from another parameter's number. This new configuration allows subscribers to explicitly manage the count of active replication origins they track.

• New GUC in PG 18 clarifies the limit on tracked replication origins since v9.5.
• Subscribers can now explicitly configure origin tracking limits.
• Resolves long-standing technical debt from a 10-year-old comment.
• Not available in PG 14-17 where the limit was implicitly tied to another param.

Logical decoding retains uncommitted transactions in memory until they commit. If the transaction size exceeds the allocated memory budget, PostgreSQL spills the data to disk. This behavior is governed by the logical_decoding_work_mem configuration parameter.

• Large uncommitted transactions can degrade performance by triggering disk I/O.
• Adjust logical_decoding_work_mem to balance memory usage and disk spill frequency.
• Monitor logical decoding activity to identify transactions exceeding memory limits.

AWS outlines a strategy for monitoring Amazon DocumentDB by categorizing CloudWatch metrics into Critical, Warning, and Advisory tiers to prioritize incident response. The guidance emphasizes diagnosing slow queries using three distinct analytical lenses to isolate performance bottlenecks. It also highlights the importance of monitoring garbage collection health proactively to prevent cluster-wide degradation.

• Organize CloudWatch metrics into Critical, Warning, and Advisory tiers to filter noise.
• Diagnose slow queries using three complementary analytical perspectives.
• Monitor garbage collection health early to avoid cluster-wide incidents.
• Shift from reactive alerting to proactive tiered signal management.

The PostGIS team has issued the second release candidate for version 3.7.0, following up on rc1 with additional bug fixes. This major release candidate supports PostgreSQL versions 14 through 19 beta 3 and requires GEOS 3.10 or higher, with full feature parity needing GEOS 3.15+. It also mandates Proj 6.1+, libgmp, and specific versions of SFCGAL and GDAL for extended functionality.

• Validates compatibility with PostgreSQL 19 beta and GEOS 3.15 for full feature sets
• Includes incremental bug fixes from the previous rc1 release candidate
• Requires GEOS 3.10+ minimum, but 3.15+ is necessary to unlock all geometry functions
• SFCGAL 2.3.0+ and GDAL 3+ are strict dependencies for their respective extensions
• Docs updated in English, Japanese, Simplified Chinese, and French

Jimmy Angelakos presented pgSafe at PGDay UK 2026, detailing the redesign of enterprise Postgres backup tools from C to Go. The talk addressed the specific rules and architectural constraints required for robust backup solutions, contrasting the new Go-based approach with established tools like pgBackRest. Audience feedback was positive, highlighting the potential of this modernized stack for managing Postgres data safety.

• pgSafe targets the enterprise backup niche, competing with established C-based tools like pgBackRest.
• Re-architecting in Go addresses modern performance and concurrency needs for backup operations.
• The tool enforces strict design rules to ensure reliability and data integrity during backup processes.
• Early reception suggests strong interest in Go-based alternatives for critical Postgres infrastructure.