How We Load Test Argo CD at Scale: 1,000 vClusters with GitOps on Kubernetes.

An interesting benchmark of ArgoCD. While the setup is somewhat too specific (vCluster), this benchmark provides some insights into the limitations of ArgoCD at scale. There are also manifests available, so you can reproduce the experiment (using some YAML generation with Bash :D)

#kubernetes #gitops #argocd

Ktea - pronounced "K-tea" - is a TUI client for Kafka that allows you to see the topics and consumer groups, produce and consume messages, etc.

It's a bit rough and you may need to configure the Meta button for your terminal (can be a bit of a PITA on MacOS), but otherwise it may be useful!

Another way of working with Kafka in more or less visual way is kafka-ui from Provectus. It's a web GUI, but it can be spun up locally in Docker.

#kafka

​​​​For today’s Donations Monday, I’d like to remind you about the fundraiser that DOU has started for the 3rd Assault Brigade.

They are raising 20M UAH (~€410k) for reconnaissance drones for the 3rd Assault Brigade.

Here's the direct link to the Monobank Jar:
https://send.monobank.ua/jar/AGK8qiQwQX

There is also a raffle for donations of more than 300UAH. If you know to know more about the raffle, as well as about the fundraiser itself, check out the dedicated page (information is in Ukrainian):
https://dou.ua/triyka/

#donations #Ukraine

Under daily routine, it's easy sometimes to forget, why we've gotten into the industry in the first place.

This is where pet-projects come in handy. Also, they are great for learning! Unfortunately, the "pet-project" term got a commercialized a bit. However, this article gave me a new term - "toy software"! I'm gonna use it from now on.

There are a few examples of such toy software that the author wrote through the years. Almost all those examples look super-complex to me. However, one need to understand that each of us has different expertise, and thus different things look easy or hard to each of us. I can easily think of a couple of examples of such platform-related toy software one could build. For example, Cost Exporter really took maybe a week, if I don't account for all the procrastination time.

#programming

A new issue of the CatOps Newsletter is out!

https://newsletter.catops.dev/p/catops-digest-2025-07-03

#digest #newsletter

Security researches from WIZ have a fun capture the flag challenge for you!

It has monthly challenges and a leaderboard. Feel free to participate!

#security

​​My wife is helping to raise funds for the repair shop for the "Omega Wings" unit. They have a unit that takes care of ground drones, and while there are the drones themselves, those require routine maintenance and occasional repairs.

So, these funds are aimed for the equipment for their repair shop: both tools and materials.

The main goal is 300k UAH, this helper jar is 20k.

https://send.monobank.ua/jar/5XD1eci1Ac

I would greatly appreciate your help! 2/3 of the helper jar is already there, so we can easily close it today!

#donations #Monday

A small article about software development in the time of AI -
Writing Code Was Never The Bottleneck.

As the name suggests, this article is about the fact that although many vendors try to sell their AI coding tools as a "replacement for the developers", the blockers that those tools remove were never the biggest ones.

There are companies that understand that and encourage their employees to do "assisted" development, which in turn can yield great results, if people are not afraid of being fired. What a surprise...

#ai #programming

Continuing the topic of AI: CTOs Reveal How AI Changed Software Developer Hiring in 2025 is a collection of testimonies from C-levels and technical leaders on what they pay attention to when hiring people in the AI-age.

There are many interesting and important points there. It's true, that writing code was never the biggest issue. Understanding what you're doing or what the code produced by AI does is crucial today. Especially, when you're hiring people. I don't know about you, but we already have a couple of cases, when candidates were seemingly using an AI assistant at interviews.

One interesting take away is the evolution of the coding challenge. It's not stated in this article explicitly, but we could all agree that take-home assignments are dead. However, some respondents are using code reviews as a challenge: they ask candidates to review the AI-generated code for bugs and possible improvements. This is a very interesting approach that could reveal, how people think about a problem.

So yeah, these two quotes summarize this article quite good:

 everyone's debating whether AI will replace developers, the people actually hiring them are looking for the opposite of what you'd expect.
 They don't want prompt engineers or AI evangelists. They want the developers who can clean up the mess AI creates.

and

 you're worried about AI taking your job, stop.

#ai #hiring #culture

LinkedIn is at it again!

Apparently, the Kafka's creators have overgrown its capabilities, so they decided to create a replacement called Northguard.

Northguard uses different protocol, replicates the data in the different way, and has different client-side. So, it's not an evolution of Kafka - it's its replacement.

Still, Northguard is not open source yet, so we do not have all the information to evaluate it - only some rumors from within LinkedIn. They do plan to open source it eventually, though.

Would it replace Kafka outside LinkedIn as well? I doubt it, Kafka (and more importantly its protocol) has too wide adoption to be thrown away that easily. But still, if we, as humanity, never reinvented the wheel, we would still travel in wooden carts of something...

#data #kafka

You probably heard of Figma's AWS bill already. If not, tldr is that they've listed for IPO, which means that they had to disclose their financial information. One of the findings there was that Figma spends roughly $300k a day on AWS, and many people freaked out.

In this article, Corey Quinn puts these spendings into perspective: this astonishing sum is just about 12% of their rolling revenue or $0.7 per active user, which is not that bad.

It's a short, funny read, since Corey has his own style of presenting his opinions:

 HackerNews commenters claimed they could cut Figma’s costs by “at least 30%, often more than half.” Sure, Steven; that seems credible. I’m certain your experience running a Minecraft server uniquely qualifies you to architect infrastructure for 95% of Fortune 500 companies.

#aws #money

​​Today we continue raising funds for the workshop that repairs ground drones.

2/3 of the goal is already reached. Let's push it towards the completion!

https://send.monobank.ua/jar/2nxQiPG4LU

#donations #Ukraine

A lot of people dislike Helm, and yet those are the same people, who use it. Rephrasing Bjarne Stroustrup: there are tools that people hate, and tools that no one uses.

Here's yet another example of a fair criticism of Helm.

I think the problem raised by this article lays however deeper that it appears at the first glance. Sure, it is easier to "validate" the plain YAML with your eyes than YAML with GoTemplate logic on top of it. Still, this problem is kind of solved in the "conventional" software engineering already. Normally, you do not reimplement everything from scratch, and you do not read the whole library codebase when doing your import foo. There are downsides, ofc, but somehow this whole industry works.

In my opinion, we should start treating other people's charts (modules, roles, recipes, what have you) as libraries, not "someone else's shell scripts". Thus, tests and documentation shouldn't be an afterthought, even when it comes to the internal things. Moreover, it's on you to write acceptance tests for your use-case. Maintaining some tests is much easier than parsing walls of YAML by yourself, trust me.

P.S. If you're interested in testing of Helm charts (including 3rd-party ones), I have a two-parter on this: Part I and Part II.

P.P.S. I also recall a quote from Kelsey Hightower that YAML is just Kubernetes' Assembly. The fact that you can read and understand it doesn't mean you have to.

#kubernetes #helm

By the popular vote, we are doing digests on Friday now. So, here it is!

https://newsletter.catops.dev/p/catops-digest-2025-07-18

#digest #newsletter

​​For today’s Donations Monday, I would like to remind you that the fundraiser for the ground drones workshop is still open!

https://send.monobank.ua/jar/2nxQiPG4LU

#donations #Ukraine

3 Humble bundles in one post!

So, I don’t need to make a post for each of them.

- SQL and Databases by O’Reilly.
- Learn Programming by Zenva
- Linux courses by Packt.

Cheers!

#books #bundle

Supercharging Development With Dapr and mirrord.

I want to share this article because of the latter. Mirrord looks very similar to Telepresence - it can mirror traffic from an existing cluster to your local setup instead of spawning things in the cluster itself. It would be great to see a comparison between the two, though.

Dapr, on another hand, is kind of a more niche project. It can abstract different brokers behind a unified API. I first learned about it at CfgMgmt Camp a couple of years ago. The problem is that if you do need to abstract several brokers in the scope of a single codebase, that must be a special case. Also, since you need to keep the API universal, Dapr only provides as many features as the least feature-rich broker does, which is a natural limitation.

Anyway, if you have used Dapr in production, that would be interesting to hear your insights!

#kubernetes

Today, I'd like to share two articles on Aurora with you:

- One on how Atlassian migrated their RDS instances for Jira to Aurora
- The second is on the blue/green upgrades functionality in Aurora


#aws #databases #db

​​For today's Donations Monday, I'd like to remind you about a big fundraiser by DOU UA for the 3rd Assault Brigade, which has recently crossed its 50% milestone!

https://send.monobank.ua/jar/AGK8qiQwQX

More information is available on a dedicated web page (in Ukrainian).

#donations #Monday

You likely know about it already, but to reiterate.

Bitnami introduces changes to its catalog from August 28th, 2025.

tl;dr: All the existing images will be moved to the legacy repository, and only the latest tag would be available for free from now on. If you want to continue using their mainline images, you have to pay a subscription fee.

Open source Helm charts will continue being open source, but "enterprise-grade" (whatever that means) charts would be available for subscription as well.

So, check your images & charts!

#kubernetes #docker #bitnami

"Invert, always invert", - Friedrich Heinrich Jacobi, probably.

There is a ton of articles online on how to make your database faster, but how many articles are there about how to make your database slower?

Making Postgres 42,000x slower because I am unemployed is a fun read about how to almost grind Postgres to a halt by only tweaking the postgresql.conf. No specially crafted slow queries - only config.

As the result, the author managed to make his Postgres installation 42 000x times slower compared to the initial benchmarks. tl;dr: tinkering around with the number of IO operations (cache size, vacuum periods, WAL checkpoints) and making the IO single-threaded do the trick here.

#databases #postgresql