OffNet Newsroom

Daily topic roundup

Database Technology

Saturday, October 03, 2026 · 8 stories, curated & summarized — click any story for the source.

Share this roundup on LinkedIn Full edition
AWS What's New awsdatabase

Aurora DSQL introduces partial indexes for targeted query performance

Amazon Aurora DSQL now supports partial indexes, allowing you to index only a specific subset of rows defined by a WHERE clause. This approach keeps index storage costs low and improves query performance by limiting data reads to qualifying rows, such as active records in a table dominated by historical data. The feature automatically applies to queries whose filters match the index condition and is available in all supported regions.

  • Index only active or high-value rows to reduce storage overhead and I/O.
  • Ideal for tables with large historical datasets and small working sets.
  • Queries matching the index condition automatically leverage the partial index.
  • Available in all regions where Aurora DSQL is currently supported.
CHECKLISTOptimizing With Partial IndexesIndex only active or high-value rowsIdeal for tables with large historical datasetsQueries matching condition automatically leverage the indexAvailable in all supported regions
Read at source →
Planet PostgreSQL database

PgBouncer 1.26.0 Fixes Two Unauthenticated CVEs

PgBouncer version 1.26.0 has been released to address three distinct security vulnerabilities. Two of these issues can be triggered without any client authentication, posing a significant risk to exposed poolers. The update is critical for maintaining the integrity of connection pooling infrastructure.

  • Two CVEs are exploitable without authentication, requiring immediate attention.
  • Upgrade all PgBouncer instances to version 1.26.0 as soon as possible.
  • Verify exposure of pooler ports to ensure unauthenticated access is blocked.
CHECKLISTPgBouncer 1.26.0 Response PlanPatch all instances to 1.26.0 immediatelyVerify port exposure for unauthenticated accessPrioritize fixes for unauthenticated CVEs
Read at source →
Planet PostgreSQL database

PostgreSQL GUCs: How min_parallel_scan_size shapes planner decisions

Christophe Pettus explains that min_parallel_index_scan_size and min_parallel_table_scan_size act as the minimum threshold for considering parallel scans. These settings also influence the planner's scaling logic, effectively setting the baseline for how cost estimates rise as scan sizes increase. Understanding this relationship is key to tuning parallel query performance for various table sizes.

  • Parallel scans require table/index size to exceed min_parallel_*_scan_size thresholds.
  • These GUCs set the baseline for the planner's cost scaling ladder.
  • Adjusting these values impacts how the planner evaluates small vs. large scans.
  • Review these settings if parallel plans are unexpectedly chosen or skipped.
Read at source →
Planet PostgreSQL database

PostgreSQL 17 Job Queues: Noisy Neighbors and Planner Traps

Chris van Eijk benchmarks tenant isolation in PostgreSQL 17 job queues, contrasting FIFO against round-robin strategies using SKIP LOCKED. The study highlights how a single tenant submitting 1,000 jobs can starve ten others submitting five each. It also identifies a specific 10-second planner trap that can distort performance expectations during testing.

  • SKIP LOCKED is the core mechanism for fair job distribution in PG 17.
  • FIFO strategies risk starvation when one tenant dominates queue volume.
  • Round-robin offers better isolation but requires careful implementation.
  • Watch for a 10-second planner trap that skews benchmark results.
  • Monitor per-tenant job counts to prevent noisy neighbor effects.
Read at source →
AWS What's New awsdatabase

Amazon RDS for MySQL Preview Now Supports MySQL 26.7

Amazon RDS for MySQL 26.7 is now available in the Database Preview Environment, marking the first Innovation release to use the new YY.M calendar versioning. This pre-release includes the Change Stream Applier for improved replication throughput and backlog recovery, along with standard bug fixes and security patches. Instances in this preview are retained for up to 60 days and snapshots are restricted to the preview environment.

  • MySQL 26.7 adopts YY.M versioning, signaling a shift in release cadence tracking.
  • New Change Stream Applier boosts replication apply throughput and recovery speed.
  • Preview instances auto-delete after 60 days; snapshots cannot leave the preview.
  • Test workloads should validate the new applier before general availability.

Fleet impact: Evaluate MySQL 26.7 in non-production RDS instances to assess the Change Stream Applier's impact on replication lag and throughput. Plan upgrade paths for your MySQL fleets once the version moves out of preview, noting the new YY.M versioning convention.

BY THE NUMBERSThe headline number60daysAmazon RDS for MySQL Preview Now Supports MySQL 26.7
Read at source →
Planet PostgreSQL database

ParadeDB Boosts Text Search via PlanetScale-Inspired BM25 Optimizations

ParadeDB has accelerated its text search capabilities by implementing two specific BM25 optimizations. These changes were directly inspired by benchmark configurations from PlanetScale's recent TIN benchmarks, allowing for significant performance gains. The update improves speed without requiring any modifications to existing document identifiers.

  • BM25 optimizations derive from PlanetScale's TIN benchmark methodologies
  • Performance gains achieved without altering current document ID structures
  • Direct integration of external benchmark insights into ParadeDB engine
  • Validates cross-platform performance tuning strategies for search engines
Read at source →
Percona Blog database

Percona ClusterSync for MongoDB Adds Active-Standby HA Failover

Percona has upgraded ClusterSync for MongoDB to support active-standby high availability, addressing the previous single-point-of-failure risk. This enhancement ensures continuous change-stream replication during long-running migrations that involve initial clones and sustained data sync. The update aims to provide greater reliability for critical infrastructure during complex MongoDB cluster migrations.

  • Eliminates single-point-of-failure in Percona ClusterSync for MongoDB
  • Enables active-standby failover for sustained change-stream replication
  • Reduces risk during long-duration migrations with initial clones
  • Improves infrastructure reliability for critical MongoDB sync tasks
TRADE-OFFClusterSync UpgradePrevious SetupSingle point of failureRisk during long migrationsReplication pauses on failureActive-Standby ModeContinuous change-stream replicationAutomatic failover capabilityCritical infrastructure reliabilityvs
Read at source →
Planet PostgreSQL database

pg_clickhouse v0.11.0 and chdb v0.1.2: Encoding fixes, nesting, and types

The latest releases of pg_clickhouse and chdb enhance cross-database compatibility through updates to their underlying header-only C libraries. A key fix addresses character encoding validation, raising exceptions for invalid bytes in text or JSON columns to prevent data corruption. These updates also introduce improvements to type handling and nesting capabilities for better interoperability between PostgreSQL and ClickHouse.

  • Encoding validation now blocks invalid bytes in text/JSON columns, preventing silent corruption.
  • Updates rely on patched clickhouse-c and pg-clickhouse-c header-only C libraries.
  • Cross-database compatibility is further strengthened with new type and nesting support.
  • Existing users should review for potential exceptions on previously invalid data.
Read at source →
↑