Zach Hurt

Stop calling your Support Engineers “SREs” (until they actually are). I’m seeing a growing trend where companies rebrand Support or Operations teams as Site Reliability Engineering (SRE) overnight. Modernizing operations is great - but changing a title doesn’t change the nature of the work. Here’s why “SRE-washing” can lead to burnout: 1 - The Mindset Gap Support is typically reactive — responding to incidents and tickets. SRE is proactive - treating reliability as an engineering problem solved with automation and software. If the team spends 100% of their time fixing issues manually, they’re still doing support. 2 - The Skill Gap SRE requires strong engineering capabilities: automation, coding, CI/CD pipelines, infrastructure-as-code, and observability. Without training and time to develop these skills, teams are set up for expectation failure. 3 - The Time Gap One of the core SRE principles is the 50% rule: no more than half of the team’s time should go to operational toil. If engineers are buried in a ticket queue with no time allocated for automation or reliability engineering, the label becomes meaningless. The real risk? When every operational role becomes “SRE,” the discipline gets diluted. Leadership expects engineering outcomes, while teams are still structured for reactive support. True SRE isn’t a title change — it’s an operating model change. Automation. Error budgets. Engineering time for reliability. Build those first. The title will follow. Is SRE becoming the new “DevOps” - a term everyone uses but defines differently? Curious to hear how others are handling this transition. #SRE #SiteReliabilityEngineering #DevOps #EngineeringLeadership #PlatformEngineering #Reliability

4

1 Comment

The bottleneck in the development cycle will almost certainly be humans. Not just because humans stink at parallelizing tasks, but because I suspect some leaders will not adapt quickly enough to the kaleidoscope of changes. In my opinion, just about any decent developer/programmer armed with some of these coding tools and an imagination is a force right now. I’ve learned frameworks I used to feel would be too much of a burden to learn and built real solutions with them. Also, starting new projects doesn’t feel like a mountain, full of stack-overflow searches and confusing tracebacks, that I have to climb anymore but, rather, a cool learning experience. And if you are a developer concerned about spending 100 bucks a month, you probably need to consider where your industry is headed and maybe give it a shot.

3

Week 2 of Spring into AI challenged us to write an tool that AI can use. I decided to create a SQL Server DB Profiler skill that an agent can invoke. I wanted to move metrics closer, within my IDE. Usage: "DTUs comin in hot! Check last 3 hours. Show top 3 offenders." Output: A synopsis of top N offending queries over last N hours, invocation counts, cpu waits, and tuning recommendations. I chatted with GPT about what kind of metrics are possible to query SQL Server for. Then I asked it to produce a small suite of sql scripts and a powershell script to map input sentiment to input params in the sql scripts, and aggregate the output into a synopsis that relays the real current state of things, so I can quickly take action. Great for when those monitoring alarms go off.. I mean, of course, they never do, but just in case ;) src: https://lnkd.in/gnRDjsNd https://lnkd.in/gFPY2sAn

9

You provisioned your DynamoDB table for 10,000 WCU. That was for a product launch. Two years ago. Current actual usage: 400 WCU. You’re paying 25x what you need. DynamoDB provisioned capacity is dangerous because it’s invisible. You set it once, maybe during a traffic spike or product launch. Then you forget about it. The table keeps running at that capacity forever. Even when traffic drops to nothing. At $0.00065 per WCU/hour: 10,000 WCU = $4.68/hour = $3,369/month Actual need of 500 WCU = $0.23/hour = $168/month That’s $3,200/month in waste. $38,400/year. The fix takes 2 minutes: Option 1: Switch to On-Demand pricing Pay per request. No capacity planning. Great for unpredictable workloads. Option 2: Enable Auto-Scaling Set a min and max. Let DynamoDB adjust based on actual traffic. Option 3: Right-size your provisioned capacity If your traffic is predictable, just lower the numbers. Check your tables right now: aws dynamodb describe-table –table-name YourTable –query ‘Table.ProvisionedThroughput’ Then look at CloudWatch metrics for ConsumedWriteCapacityUnits. If consumed is 5% of provisioned, you’ve got a problem. How long have your DynamoDB tables been running at launch-day capacity?

5

2 Comments

I recently rewatched Inglourious Basterds, and the "I speak the most Italian" scene got me thinking about how Redis and other databases offer support for vector workloads. You know the one, where Brad Pitt says "Buongiorno," like he's ordering BBQ in Tennessee. 🍖 Now imagine that scene—but instead of spies trying to sneak into a theater in disguise, it’s databases pretending to be vector stores. 😏 🐘 #PostgreSQL (Enzo Gorlomi) Postgres walks in with its new vector extensions. "Buongiorno!" he says. Technically correct… but you can still hear the accent. Those B-trees are trying their best to pronounce "cosine similarity." 🍃 #MongoDB (Antonio Margheriti) MongoDB joins in confidently. "Antony Marge-retty," he says. It's flexible, but you can tell it's thinking in JSON, not in vectors. 🍝 #Redis (Dominic DeCocco) Then Redis enters. Calm, smiling, confident, yet humble. It doesn't pretend to understand vectors. It speaks them. Natively. Effortlessly. Perfect accent. With sub-millisecond (or ultra-low) latency. ⚡ Moral of the story? Some databases may offer you vector features. Redis? It gives you the same—but with fluency. No staggering. No hiccups. No long pauses. And, fun fact… it was built by someone who actually speaks Italian. 🇮🇹 😅

224

7 Comments

Finish this sentence: Advising teams who do not write tests when developing software to adopt Trunk Based Development is akin to ____________________

1

3 Comments

Anyone who uses Claude Code is now an engineering manager. And as Jocko Willink explains: “Every problem is a leadership problem.” Those of us who now manage the new “managers of agents” have additional responsibilities to ensure that effective guidelines, procedures, tools, and training are in place. Our race to push through 10x delivery records can be a “sugar high” with a painful hangover.

48

What makes a good tool? The The Pragmatic Engineer released their 2025 survey of over 3k software engineers.  Part of the survey was around most disliked tools that engineers have to use Engineers are lazy (and they should be lazy).  They dislike tools that are inefficient, cause repetitive work, or are hard to use.  They especially dislike tools that do too many things, and do the most obvious, important things poorly. 💭 Food for thought

1,091

520 Comments

Here's the thing about MQ vs Kafka that nobody talks about enough: They're "NOT" interchangeable. I learned this the hard way three years ago when our team tried using Kafka for something that should've been a straightforward MQ job. We spent weeks fighting with consumer groups and offset management when all we needed was reliable point-to-point messaging. 🤦‍♂️ So when should you actually use what? Use Message Queues (RabbitMQ, IBM MQ, ActiveMQ) when: ✅ You need guaranteed message delivery with acknowledgments ✅ Point-to-point or request-reply patterns are your primary use case ✅ Message priority matters (urgent messages first) ✅ You're dealing with transactional workflows Use Kafka when: ✅ You're building event streaming architectures ✅ Multiple consumers need to read the same data independently ✅ You need message replay (going back in time) ✅ High throughput is critical (millions of messages/sec) ✅ You're doing real-time analytics or event sourcing The honest trade-offs: ✅ MQ Pros: Simpler setup, better for complex routing, mature protocols ❌ MQ Cons: Harder to scale horizontally, messages are deleted after consumption ✅ Kafka Pros: Insane throughput, message persistence, built for distributed systems ❌ Kafka Cons: Steeper learning curve, overkill for simple use cases, more operational overhead Here's my rule of thumb: If you're asking "do I need Kafka?" you probably don't. Start with MQ. But if you're building something event-driven at scale? Kafka's your friend. What's been your experience? Ever regretted choosing one over the other? Follow me on LinkedIn: https://lnkd.in/dkeKb6cc #Kafka #MessageQueue #SystemDesign #SoftwareArchitecture #EventDrivenArchitecture #Microservices #DistributedSystems #TechLeadership #EngineeringBestPractices #ThePracticalArchitect

23

4 Comments

"It’s only NPE (non-prod environment), so it’s not urgent. I’ll look at it when I can." As a software developer, NPE is our production. Every new feature, bug fix, and business-critical change starts in non-prod. If NPEs are broken or unreliable, work gets delayed, timelines slip, and ultimately production does get impacted, just indirectly. So next time someone flags an issue in NPE, please consider this: Fixing NPE issues quickly keeps delivery moving and ensures production stays healthy.

19

2 Comments

Accountability is the great equalizer of leadership. Without it, standards slip. Trust erodes. Culture drifts. I’ve led teams from 12 to 200+ people across commands spanning 100,000 square miles - and the most consistent factor in every high-performing team I’ve seen is accountability. It’s what aligns people behind one standard, regardless of rank or personality. In my latest video, I break down: 🔹 Why accountability separates good leaders from great ones 🔹 How to correct “organizational drift” before it destroys culture 🔹 When to move from verbal to written accountability 🔹 Why accountability can be both corrective and caring Watch the full breakdown here: 🎥 https://lnkd.in/eT2W6eZZ Let’s start a conversation - how do you build accountability without crushing morale? #Leadership #Accountability #OrganizationalCulture #LeadershipDevelopment #TeamBuilding #MilitaryLeadership #ElderMasterChief

1