Newsdesk
Engineering
Data & AI
Industries
Enterprise Systems
Go-to-Market
Longform
DesignIndiaAll stories

Databases & Data

27 October 2026

75 new posts across 9 database and data engineering sources — 9 worth your time.

One week of Databases & Data, 9 stories, as published.

Reliability

Jeremy Schneider: Postgres recovery_prefetch Considered Harmful and Causing Outages

Jeremy Schneider highlights real-world production outages caused by PostgreSQL's recovery_prefetch setting. The post explores how background prefetching during recovery can quietly introduce unexpected issues that disrupt database availability.

Why it matters — Audit default recovery flags like recovery_prefetch, as aggressive background prefetching during WAL replay can lead to unpredicted production deadlocks or outages.

Planet PostgreSQL · 27 October 2026 · Read the original →

Postgres

Shard design doesn't start from your tables

PlanetScale explains how sharding child tables by parent entity keys prevents cross-shard router joins. By mapping orders using customer_id in a single shard group, joined queries land cleanly within individual Postgres instances.

Why it matters — Co-locating related records using a shared partition key avoids expensive cross-shard joins in sharded relational architectures.

PlanetScale · 27 October 2026 · Read the original →

Analytics & OLAP

Front-end observability at scale: How OpenReplay captures billions of daily events in real time with ClickHouse

OpenReplay replaced Postgres with ClickHouse to handle high-throughput telemetry ingestion, scaling up to 6 TB per day on a single vertically scaled machine. The transition enabled sub-second analytical queries and automated retention management through TTL settings.

Why it matters — Replacing transactional databases with a columnar engine for log ingestion can deliver sub-second queries on 6 TB/day on a single node.

ClickHouse · 27 October 2026 · Read the original →

Reliability

Christophe Pettus: All Your GUCs in a Row: recovery_min_apply_delay

Christophe Pettus examines PostgreSQL's recovery_min_apply_delay, a setting that forces a read replica to lag behind the primary intentionally. This artificial delay creates a queryable safety copy where accidental data deletions can be observed and corrected before replay occurs.

Why it matters — Intentionally delaying WAL application on a dedicated standby provides a window to salvage data after destructive queries without restoring full backups.

Planet PostgreSQL · 27 October 2026 · Read the original →

Data Engineering

How Obie Insurance cut its month-end close from 25 hours to 2 with dbt State and a lake house rebuild

Obie Insurance overhauled its data pipeline architecture by migrating to Apache Iceberg and using dbt State. The combined Lakehouse rebuild cut its critical month-end data processing job from over 25 hours down to 2 hours.

Why it matters — Combining stateful incremental execution with modern open table formats reduced pipeline runtime by over 90%.

dbt · 27 October 2026 · Read the original →

Postgres

Ahsan Hadi: Zero Downtime Rolling Major PostgreSQL Upgrades on a Distributed PostgreSQL Cluster

Ahsan Hadi discusses performing zero-downtime major version upgrades on PostgreSQL using Spock multi-master logical replication. Rolling upgrades allow nodes to be updated sequentially while the cluster continuously serves application traffic.

Why it matters — Multi-master logical replication allows major database upgrades node-by-node without taking application downtime.

Planet PostgreSQL · 27 October 2026 · Read the original →

Practice

I don't trust ON DELETE CASCADE

PlanetScale details the hidden operational risks of using ON DELETE CASCADE foreign key constraints in high-scale environments. Automatic cascading deletions can trigger unintended locks and long-running transactions across dependent tables.

Why it matters — Avoid relying on cascading constraints for large relational tables; explicit batching in application logic prevents lock escalation and write spikes.

PlanetScale · 27 October 2026 · Read the original →

Reliability

ClickHouse Managed Postgres Expands Notifications

ClickHouse Managed Postgres now provides expanded customer notifications for critical operational events. These include alerts for inactive replication slots (triggering after 10 GiB of unconfirmed WAL over two hours), planned switchovers, unplanned failovers, long-running transactions, and connections nearing their limits.

Why it matters — Database owners can proactively address potential issues like replication lag and connection exhaustion, improving the reliability and operational visibility of their ClickHouse Managed Postgres instances.

ClickHouse · 27 October 2026 · Read the original →

Performance

Shipsidekick: 10x Faster Queries with ClickHouse Postgres

Shipsidekick migrated its database from Neon to ClickHouse Managed Postgres, achieving a 10x improvement in query performance and enhanced reliability. This user story highlights the benefits of ClickHouse's managed PostgreSQL offering for applications requiring faster data retrieval.

Why it matters — Teams using Neon for PostgreSQL may find ClickHouse Managed Postgres offers a significant performance advantage for their query-intensive workloads, potentially justifying a migration.

ClickHouse · 27 October 2026 · Read the original →

The industry remains divided between those who trust their constraints and those who have used them.

9 stories, every Tuesday

Published here every week. Follow by RSS to get it as it lands.

← Previous issue Next issue →