Command Prompt, Inc.

09/01/2026

When does solving a PostgreSQL problem create another one?

Ildefonso Camargo looks at a few places where that tradeoff shows up: LDAP bind credentials in pg_hba.conf, the added complexity of PAM and Kerberos, and large objects that logical replication still doesn’t support.

Watch the full T3D Postgres Wishlist conversation: https://youtu.be/YMcW7iWgYQ8

08/31/2026

Have you ever looked back at a stressful situation and realized you made it bigger in your head than it actually was?

Amanda Nystrom and JD discuss why our first reaction isn't always the full story, and how changing the way we look at a situation can change the way we experience it.

Watch the full episode here:

In this episode of Health and Tech, Amanda Nystrom and JD Drake exp...

08/27/2026

What becomes more valuable when AI makes average work easier to produce?

Judgment. Experience. The ability to see what others miss.

In episode 59 of More Than a Refresh, Nikki Barua talks about why the next stage of AI is not only about learning new tools. It is also about understanding what you uniquely bring to the work, then using AI to extend it.

The conversation moves from identity and accountability to what it means to become a builder again, especially for leaders who have spent years moving further away from ex*****on.

Watch More Than a Refresh 59: Unlocking Your True Gift: https://youtu.be/nLdoHk68tjo

AI is changing more than how work gets done. It is changing what be...

08/25/2026

For a national employer benefits administrator (TPA) client, database recovery issues quickly escalate into compliance risks.
When an urgent recovery incident hit, we resolved the immediate bottleneck and guided them through a systemic audit of their autovacuum settings, Change Data Capture (CDC) replication lag, and transaction logging.
True resilience isn't about a single configuration setting; it's about engineering an environment for continuous compliance and structural stability.
Let's discuss securing your PostgreSQL database:
https://www.commandprompt.com/contact-us/

08/20/2026

In Part 1 of Beyond the 25% Rule, Greg Dostatni looked at how different workloads respond to changes in shared_buffers.
Part 2 moves to huge_pages.
Greg repeated the benchmarks using 2 MiB and 1 GiB huge pages. In these tests, neither produced a significant performance improvement. The workloads were constrained elsewhere, so reducing memory translation overhead did not change throughput much.
What did change was page-table memory use. With standard 4 KB pages, the kernel used as much as 20 GB in some of these tests to keep track of memory addresses.
Greg then uses Linux PSI to look at where the system is actually under pressure before deciding whether huge_pages are likely to help.
Part 2 of Beyond the 25% Rule: When huge_pages Help: Using PSI to Find Bottlenecks Before You Decide
Part 2:

Benchmarking huge_pages on a 128 GB PostgreSQL 18 system shows throughput gains are modest — but page table overhead at 4KB pages can consume up to 20% of total RAM. Use PSI to find where your system is actually stalling before making the call. Part 2 of 3.

08/19/2026

AI doesn't replace engineering judgment.
Command Prompt DBA Brian Fehrle shares why keeping humans in the decision loop still matters when AI becomes part of your workflow.

Watch the full Group Swim discussion: https://youtu.be/srFsNtoCLF4

08/18/2026

Correct code isn’t always fast code.

Our Mini Cython Series is now available on YouTube. Four sessions on cutting through Python overhead with Cython, compiling and benchmarking your code, and using typed memoryviews to speed up numeric loops when they become your bottleneck.

The full series is here:

Learn how Cython can speed up Python code without requiring you to rewrite everything in C. In this Mini Cython Series, we start with the fundamentals and bu...

08/17/2026

Most of us spend a lot of time talking to ourselves.

The question is whether those conversations are actually helping.

Amanda Nystrom and JD Drake explore how judgment, curiosity, and the questions we ask ourselves can influence the way we experience stress.

Listen to the full episode:

In this episode of Health and Tech, Amanda Nystrom and JD Drake exp...

08/13/2026

We’re starting a new three-part blog series on PostgreSQL shared_buffers behavior in large-memory systems.
The usual starting point is 25% of system memory. Greg Dostatni tested how well that guidance holds on a 128 GB system, varying shared_buffers from 1 GB to roughly 110 GB across two very different workloads.
The TPC-B-style point-update and analytics workloads did not tell the same story. In the point tests, 25% was a reasonable setting. In the analytics tests, it was well away from the strongest-performing ranges.
That is the problem with treating a starting point as a rule. The workload, working set, and balance between PostgreSQL and the OS file cache all change the result.
Part 1 of Beyond the 25% Rule covers the benchmark setup and the initial results.
Learn what he discovered here:

Tuning shared_buffers isn't just about peak throughput — it's about how much CPU and disk I/O a system consumes per unit of work. The 25% default can require up to 5x more disk I/O than very low or very high allocation. Part 1 of 3.

08/12/2026

PostgreSQL systems are shaped by more than the database itself.
Past decisions, changing workloads, and the infrastructure around PostgreSQL all affect how an environment behaves in production. That is where Command Prompt begins.
We provide PostgreSQL consulting and hands-on engineering. We can review an important technical decision, investigate a problem that has resisted the usual fixes, or work directly with an internal team as changes move into production.
When the issue extends beyond PostgreSQL, we look at the systems around it too.
https://www.commandprompt.com/contact-us/