Shana Kempfer’s Post

🚀 Let’s take a quick turn from replication and talk about something just as important in PostgreSQL environments: connection pooling. As systems grow, one challenge shows up fast: Too many connections. PostgreSQL handles connections using a process-per-connection model, which is powerful but can become expensive when hundreds or thousands of clients connect at the same time. That’s where connection pooling comes in. Instead of every client opening a direct connection to the database, a pooler manages and reuses connections efficiently. A few common approaches stand out. 🔹 pgBouncer Lightweight and very fast. It focuses purely on connection pooling and is often used in high-throughput systems. Great for reducing connection overhead. 🔹 pgpool-II More feature-rich. In addition to pooling, it can handle load balancing, query caching, and even failover management. Useful in more complex setups. 🔹 RDS Proxy (managed option) For AWS environments, RDS Proxy provides managed connection pooling. It helps stabilize connections, improves scalability, and reduces failover impact without needing to manage the pooler yourself. Why does this matter? Because without pooling, too many connections can lead to: • memory pressure • CPU overhead • slower query performance • connection storms during spikes With pooling, you get: • better resource utilization • improved scalability • more stable database performance • smoother failover behavior In many production environments, connection pooling becomes a must-have layer between applications and the database. Because sometimes, improving performance isn’t about tuning queries. It’s about managing how connections are handled. #PostgreSQL #ConnectionPooling #pgBouncer #pgpool #RDS #DBA #DatabaseArchitecture #Performance

🚀 🚨🔥After exploring backup and recovery across databases, I want to shift the focus to something just as critical in PostgreSQL: High Availability (HA). Because backups help you recover… But HA helps you avoid downtime in the first place. PostgreSQL has evolved significantly in this space, and today it offers a very solid and flexible HA architecture. At the core, PostgreSQL supports two main replication models: Physical Replication This operates at the block level. Changes from the primary are streamed via WAL (Write-Ahead Logs) to standby servers. It’s fast, efficient, and commonly used for failover and read replicas. Logical Replication This works at the data level. Instead of copying blocks, it replicates tables and changes logically. It’s more flexible and is often used for selective replication, migrations, or cross-version upgrades. In a typical high availability setup, you’ll find: • a Primary node handling writes • one or more Standby replicas receiving WAL changes • a replication mechanism (streaming or logical) • a failover strategy using tools like Patroni, repmgr, or cloud-managed services • optional load balancing for read traffic When designed properly, this setup provides: • minimal downtime during failures • automatic or controlled failover • read scalability • strong data consistency What makes PostgreSQL stand out is how transparent and flexible its HA architecture is. You’re not locked into a single approach. You can design HA across on-prem, cloud, or hybrid environments, and integrate with orchestration tools depending on your needs. Over the years, PostgreSQL has continued to improve replication performance, failover handling, and operational tooling — making it one of the most reliable open-source databases for high availability systems. Because in modern systems, it’s no longer just about storing data. It’s about making sure your data is always available when your applications need it. #PostgreSQL #HighAvailability #DatabaseArchitecture #DBA #DataPlatforms #OpenSource

🚀 𝗖𝗹𝗼𝘂𝗱 & 𝗱𝗮𝘁𝗮𝗯𝗮𝘀𝗲 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻𝘀 𝗱𝗼𝗻’𝘁 𝗵𝗮𝘃𝗲 𝘁𝗼 𝘁𝗮𝗸𝗲 𝗺𝗼𝗻𝘁𝗵𝘀 - 𝗻𝗼𝘁 𝗮𝗻𝘆𝗺𝗼𝗿𝗲. Organizations want to modernize faster, cut licensing costs, and move to open-source or cloud-native databases without the usual friction. At #HexaCluster, we’ve built a migration ecosystem designed exactly for this: 🔹 𝗗𝗠𝗔𝗧 - Automated database assessment, compatibility analysis & remediation planning in minutes. 🔹 𝗛𝗲𝘅𝗮𝗥𝗼𝗰𝗸𝗲𝘁 - Enterprise-scale migrations and replications - with granular control. 🔹 𝗛𝗲𝘅𝗮𝗕𝗿𝗶𝗱𝗴𝗲 - Extension to provide industry-leading Oracle compatibility to community PostgreSQL. Together, these tools empower DB teams to halve their migration timelines, de-risk every stage, and confidently move workloads to PostgreSQL, cloud platforms, or modern architectures. Ready to see it in action? Reach out at connect@hexacluster.ai 🚀 #HexaRocket #DMAT #PostgreSQL #DatabaseMigration #Oracle #MSSQL #CDC #DatabaseAssessment

Another piece of the MariaDB ecosystem puzzle falls into place. DBaasNow is joining the MariaDB Foundation as a Silver Sponsor — and what they are building addresses a very real gap: how teams actually operate databases once things get messy (multi-env, hybrid, mixed stacks… reality). MariaDB adoption doesn’t fail because of SQL syntax. It fails because operations become complex. DBaasNow is tackling exactly that layer — automation, governance, lifecycle management across environments — which is where MariaDB increasingly lives today. This is also what the Ecosystem Hub is about: not abstract databases, but real architectures. Looking forward to exploring concrete use cases together. Read more: https://lnkd.in/dNzB8t_S #MariaDB #OpenSource #Database #DBaaS #PlatformEngineering #DataOps

https://lnkd.in/gFBZ5R9c The evaluation revealed that most use cases were single-Region workloads, whereas others required multi-Region support that could be served by Aurora Global Database, which uses asynchronous storage-based replication, with typically less than 1 second cross-Region replication lag. This enables low-latency read operations for applications connecting from geographically distributed locations.

Nobody talks about the most expensive line of code in your entire codebase. It's not in your application layer. It's not in your API. It's this: SELECT * FROM orders WHERE DATE(created_at) = '2024-01-15' That single function wrap on a timestamp column bypasses the index on a billion-row table. Silent. Catastrophic. Completely avoidable. This is why Database Administrators still matter — even in a world of managed cloud databases, AI copilots, and "NoSQL fixes everything." The cloud handles your backups. It does not handle your schema decisions. It does not catch your N+1 queries. It does not know WHY that nullable column exists or what breaks if you change it. I've watched teams spend $40K/month on RDS instances that a single index strategy would have cut in half. I've seen migrations go from 6-hour outages to zero-downtime deployments — not because of better infrastructure, but because someone understood pg_repack and pt-online-schema-change. The DBA role didn't die. It evolved. Today's database specialist owns: → Query execution plan analysis → Schema design that survives scale → Zero-downtime migration strategy → Observability (pg_stat_statements, PMM, Datadog) → Vector DB integration for AI workloads If your team is "too small for a DBA," you don't have a DBA problem. You have a knowledge gap — and your database is quietly accumulating the debt. The most dangerous sentence in engineering: "The database handles it." Does your team have someone who truly owns database performance? Or is it everyone's problem and therefore no one's? #DatabaseAdministration #SoftwareEngineering #BackendDevelopment #PostgreSQL #TechLeadership

It’s KubeCon week 🚀 If you’re a developer, ask yourself: wouldn’t it be great if your company gave you a catalog of databases that you can provision directly from Kubernetes—just like any other service—through an Operator? That’s exactly where Nutanix Database Service (NDB) comes in. Provisioning is just the starting point. NDB goes further by automating the entire database lifecycle: 👉 patching 👉 scaling 👉 cloning (critical for dev/test environments) So you’re not just deploying databases—you’re operating them with a modern, Kubernetes-driven model. If your organization is serious about simplifying database operations for developers, it’s time to take a closer look at NDB. https://lnkd.in/dBs_R_G2

For developers building apps on containers, how are you connecting the database? Nutanix Database Service has a Kubernetes Operator that lets you self-serve database requests when you’re containerised app needs it. Nutanix are at KubeCon this week where you can find out more. Or, I’d recommend reading this blog.

I got a message at 2:17 AM. “Production is down. Login not working.” The SaaS product had around 40k users. And the entire platform was stuck because of one database query. Here’s what happened. A reporting feature was released earlier that evening. One query missed an index. Every time a user opened the dashboard, it triggered a full table scan on a table with ~18 million rows. Within 25 minutes: • CPU on the DB server → 100% • API latency → 12 seconds • Connection pool → exhausted • App → effectively unusable The scary part? Nothing was technically “broken”. The database was simply overwhelmed. By the time the team noticed, the damage was already done. After that incident we changed a few things permanently: 1. Query performance checks before every release 2. Automatic alerts on slow queries 3. Index recommendations based on query patterns 4. Early warnings when database health starts degrading 5. Visibility into CPU spikes, connection usage, and query load Most outages don’t start with servers crashing. They start with silent database issues that go unnoticed until users feel it. That’s exactly why we built HealthMyDB. It helps teams detect: • slow queries • missing indexes • connection saturation • abnormal DB load • early performance degradation Before they become 2 AM incidents. Curious about something 👇 What was the most unexpected thing that broke your production database? A bad query? A missing index? A migration gone wrong? Let’s hear the war stories. #SaaS #Database #StartupEngineering #DevOps #CloudInfrastructure #PostgreSQL #MySQL #AWS

We’re pleased to welcome DBaasNow as a Silver Sponsor of the MariaDB Foundation. DBaasNow is building a database operations control plane designed to simplify and standardize how organizations manage databases across hybrid and on-premise environments. As more teams run MariaDB alongside other systems, the need for a unified operational layer becomes increasingly important. Their focus on automation, governance, and lifecycle management aligns closely with one of our core priorities at the Foundation: making MariaDB easier to adopt, operate, and scale in real-world environments. This is exactly where the ecosystem matters — not just the database itself, but the surrounding tooling that removes friction and enables production use. We’re looking forward to working together on joint content, use cases, and practical guidance for teams running MariaDB in complex environments. Read more: https://lnkd.in/dhfzctZ9 #MariaDB #OpenSource #Database #DBaaS #DataInfrastructure #Ecosystem