Cloudflare shares how they use Terraform in production.

Their setup is quite standard: Terraform, Atlantis, Conftest (OPA). One interesting thing is that they use their in-house tool called tfstate-butler to work around the lack of encryption of the Terraform states. Although, they do not disclose the details of this tool.

Another catchy quote:

>>>
...we do this at a global scale — where a single misconfiguration can propagate across our edge in seconds and lead to unintended consequences.

Yeah... We know, Cloudflare, we know...

#terraform #iac

GitHub Actions will charge $0.002 per minute for self-hosted runners starting from the 1st of March 2026.

Obviously, you would still pay whatever you pay for your self-hosted infrastructure itself.

GitHub Actions will remain free for public repositories. For now.

#cicd #gha #microsoft

On the positive note: Docker opens access to their hardened images (DHI) to everyone, not just their enterprise customers.

DHI uses a distroless runtime and includes SBOM.

Here you can browse the whole catalog of DHI. Docker asked me to login, though, but I'm definitely not an enterprise customer :D

#docker #security

Good one. Have literally the same feedback. Cool tech but mostly useless.

https://johnjames.blog/posts/graphql-the-enterprise-honeymoon-is-over

Cold-Restart Resilience is an article on what could go wrong, when a system recovers from a total outage. Cases, covered in this article, with some tips on how to solve those:

- Circular bootstrap dependencies
- Using in-memory storage as databases
- Failures when trying to create a quorum
- Failures to fetch a remote dynamic config
- Stale data in leaderless systems

It doesn't mention cascading errors, but those are kinda famous already.

#sre #reliability #systems

​​For today’s Donations Monday, I would like to ask to help a friend of mine to get a car at the Zaporizhzhia front lines.

https://send.monobank.ua/jar/5mSFtTYUFt

This is a personal request, so you can be sure that this fundraiser is legit.

#donations #Ukraine

Monzo - a British neobank - reveals their system that grants engineers temporary elevated access.

tl;dr: They are using AWS Nitro Enclaves for this.

During my time at N26, we also had a system that served the same purpose, albeit it was designed differently.

#security

The last digest of this year is here!

https://newsletter.catops.dev/p/catops-digest-2025-12-27

With this digest been out, I'm taking some holidays. So, there will be no new posts here until the end of the year (it's not like there were many posts in the last couple of days, lol).

Also, I would really appreciate it, if you could share your thoughts about the newsletter in general. Unlike for the Telegram channel, I cannot really find a good fit for it. You can share your thoughts in the comments on Substack, in our chat (in Ukrainian), or via info@catops.dev

🎄🎄🎄 Happy holidays! 🎄🎄🎄

​​I'm back!

It always feels nice to start a new year from scratch. Unfortunately, it's often not the case, and we have to finish things that remained.

Today's fundraiser is one of those things: let's help a friend of mine to raise funds for a pickup truck for the Zaporizhzhia front lines:

https://send.monobank.ua/jar/5mSFtTYUFt

#donations #Ukraine

Starting a new year with a postmortem, eh?

There was a prolonged incident with Kafka at Honeycomb last month. Here you can find a preliminary postmortem for this incident.

"Preliminary" means that there is no root cause analysis yet, but there's already the timeline and the remediation steps.

#postmortem

Our job is all about tradeoffs.

This article describes tradeoffs of database indices.

#databases

I think, this could be a good Friday read: "When Change Outruns Us" is a tale about sustained progress.


The main point of this article is that smart companies do not push for "constant change for the sake of change", but rather adopt a more cyclic pace, when the periods of extensive work are followed by more relaxed times.

This article is particularly interesting to me, because I've just finished listening to the "Slow Productivity" book by Cal Newport. One of the principles, outlined in that book, is that one should work in their natural pace. However, a constant run is no one's natural pace. Another observation in that book, is that starting from the second half of the XX century, managers started to approximate work by "business", i.e. if you look busy, you do some work, even if in the reality, there are zero outcomes.

Many tech companies like to claim that they are "outcomes-oriented" or "value impact", but in my experience, "business" is still the approximation for work. Especially, once a company growth beyond the size, when everyone naturally knows everyone, as well as what they are doing.

#culture #mgmt

A topic on Reddit that argues for a new term - Claude Hole. This is when AI is tasked with fixing an issue, but you actually end up further away from the fix after the AI changes.

This reminded me of a completely unrelated article - I Was Kidnapped by Deutsche Bahn and All I Got Was 1.50 EUR - in that story, the author also ended up further away from their destination than at the beginning of their journey.

Regardless of your views on AI, distinguishing good from bad becomes more and more valuable these days.

#ai

I found a young blog that has some potential. There are two articles so far:

- Frameworks for understanding databases
- Sorted string tables (SST) from first principles

If you've read "Designing Data-Intensive Applications" by Martin Kleppmann and "Database Internals" by Alex Petrov, you may find these articles repeating information from those books. However, unless you actively work on building databases, you may easily forget this information.

So, such articles serve as a great reminder. I enjoy getting back to such "low level" details from time to time. These details help one to better understand the tradeoffs of the end solution, in my opinion.

#databases

Another Friday read on the Culture debt.

The core idea behind it is the same as for the technical debt, but this time it’s about the culture. As company growths, relationships within it inevitable change. This is silly to ignore these changes. Yet, if nothing is done, the culture can drift away, and then it’s much harder to fix, than refactoring a couple of services.

Also, this line strikes hard, when talking about the symptoms of the culture debt: “What you reward diverges from what you say you value.”

People are quite good in calling bullshit on someone or something. It’s often quite obvious when proclaimed values are just empty slogans. This erodes trust as nothing else.

#culture

“How much memory does a Kubernetes node use without the Kubernetes layer”? This is exactly the question From RSS to WSS: Navigating the Depths of Kubernetes Memory Metrics tries to figure out.

And the answer to this question might be harder to get to, as it seems at the first glance.

#kubernetes #systems

For today's donations Monday, I'd like to share once again the standing Monobank jar for FPV equipment.

This jar is for the unit in which a guy from my wife's hometown serves.

https://send.monobank.ua/jar/4WLw91UqFe

#donations #Monday

Wanna become a true Terraform SLOPerator?

Here is a carefully vibecoded solution by Anton Babenko. I can confirm that he checks the docs at least once during his Claude conversations, so you can be confident in the quality :)

Jokes aside, this is a cool Skill for Claude Code, which currently works better than any other official or popular alternative out there.

#terraform #ai #claude

Bring Back Ops Pride is a new article by Charity Majors on how it comes that the operational work is often seen as of lower importance, and why is it bad.

This is her answer to the comments under her another article “You Had One Job”: Why Twenty Years of DevOps Has Failed to Do it. This article has some interesting ideas, but it's a marketing material, so beware.

#ops #culture

Unfortunately, kubectl flame tool for profiling in Kubernetes wasn't updated in 4 years. It cannot even run on ARM-based machines.

But what if you need to profile something in your systems? You can use continuous profiling, if it's available in your observability stack.

Or you can use kubectl prof to do some ad-hoc profiles.

- Tool on GitHub
- Medium post

#kubernetes #performance