Adapting to AI tooling in DevOps

My company is making a hard shift towards AI and I'm struggling. I'm tasked with something I have a small amount of knowledge about, but not enough to fully implement. Before AI, my approach to challenges like this would be to iterate on it, work through the problems, and learn what's required. Using Claude has just been a terrible experience for me, and I'm wondering if I'm the problem, my approach is the problem, or if this is just what we're all doing now.

I explained what I wanted (essentially a DB migration pipeline from RDS -> k8s pod) and claude barfed out about 10 shell scripts and a handful of pod templates. Everything had long, meandering comments that made very little sense. Is my job now to just start running this code and feeding pipeline errors back into claude and not worrying about it?

I started to pick apart the comments, ask questions, and update with my own comments, with the goal of being able to understand the process in detail. In doing this, I found a bunch of things that just didn't make sense or weren't optimal. When I ask the AI agent about it, it says I'm right and generates a bunch more stuff that may or may not make sense. Interacting with it is making me feel like a crazy person. I hate the way it presents information back to me. It's not succinct, it pulls in context that is of negligible value and just muddies the waters, and it's been very frustrating and existentially upsetting.

I'm willing to hear that I'm using the wrong model, or that I didn't spend enough time on my .md files and skills, but I'm also wondering if this is a normal experience, or if my approach is just dated, and I need to get over my personal values about understanding what I'm generating and giving to customers? I'm not so-much asking about the specific task at hand, just about adapting and having realistic expectations and reducing friction. I'm also very happy to hear about people's struggles.

https://redd.it/1vklxud
@r_devops

Any devops from Perth here?

Hi there I have been working as a DevOps for past 5 years and I recently moved to Perth so I am finding harder here… moreover I am feeling like people still don’t understand DevOps what really is DevOps we can catch up so Perth DevOps hmu..

https://redd.it/1vl14gc
@r_devops

Sydney DevOps Meetup: AI Agents and Open Source Developer Platforms on Kubernetes

Sydney DevOps folks,

If you're working with DevOps, Kubernetes, SRE, platform engineering, or AI infrastructure, CNCF Cloud Native Sydney is hosting a meetup on August 19 around two open source projects.

The session will cover:

kagent — a Kubernetes-native framework for building, deploying, and managing AI agents, including areas such as MCP integration, observability, security, and agent workflows.

GitHub: https://github.com/kagent-dev/kagent

OpenChoreo — an open source developer platform for Kubernetes that brings together platform abstractions, CI/CD, GitOps, observability, Backstage, and developer self-service.

GitHub: https://github.com/openchoreo/openchoreo

The discussions should be especially relevant if you're interested in how DevOps and platform engineering workflows are evolving around Kubernetes and AI.

Location: Cuscal Limited, Barangaroo, Sydney
Date: Wednesday, August 19
Time: 5:30 PM – 8:30 PM AEST

The event is free and includes technical sessions, networking, food and drinks, and Q&A with the maintainers.

RSVP: https://www.meetup.com/cncf-sydney/events/316052564/

If you're around Sydney and working in DevOps or cloud native, come join us.

See you there!

https://redd.it/1vl8akn
@r_devops

EU cloud provider news roundup, May–Aug 2026: KVM escape patch wave, Redis→Valkey, uneven 1.36 rollout

I run [eucloudcost.com](https://eucloudcost.com/) (EU cloud price comparison, open source data). I also track what providers actually ship each month — blogs, changelogs, RSS, \~30 providers.

Latest edition, May 21 – Aug 6: [https://www.eucloudcost.com/blog/eu-cloud-news-may-aug-2026/](https://www.eucloudcost.com/blog/eu-cloud-news-may-aug-2026/)

Highlights:

* Januscape (CVE-2026-53359) was a KVM escape at the **hypervisor** layer — managed Kubernetes did not absorb it for you. OVHcloud patched tens of thousands of machines in a week.
* Redis is being replaced by Valkey, STACKIT deprecated Redis the same day it launched its Valkey KV store, IONOS took Valkey-backed In-Memory DB v2 to GA.
* K8s 1.36 spread unevenly: AKS shipped it GA *with LTS*, Scaleway Jul 7 - STACKIT SKE is still on 1.34, no EKS announcement at all.
* OVHcloud is contesting criminal charges in Canada over a production order. A European provider fighting a foreign lawful-access order is a live test of the whole sovereignty pitch \^\^
* Also OVHcloud: replacing Ceph with Exten, an in-house NVMe block storage engine.
* No new EU price hikes this quarter; Exoscale cut A40 GPUs 30%.
* IONOS kills Llama 3.1 405B on Sep 15 — HTTP 400, no fallback. Pinned model IDs have expiry dates now.

Covers hyperscalers too, plus a deadline table.

https://redd.it/1vld5jy
@r_devops

The unholy trinity of LLM API errors (and my dirty fix to stop agents from crashing at 2 AM)
https://redd.it/1vlgjgg
@r_devops

We Audited 156,808 Failed CI Runs Across Eight Open Source Repos
https://www.latchkey.dev/blog/we-audited-156808-failed-ci-runs

https://redd.it/1vlmbu3
@r_devops

Je suis débutant et cherche à réaliser ce projet

Bonjour à tous.

Je cherche a créer un site web ou un logiciel permettant de la saisie de donnees sur un fichier excel (ou google sheet) de manière dynamique (en utilisant BEAUCOUP de questions soumises à conditions) et si possible d'une loop.

Le fichier excel est déjà existant et il a été crée comme un questionnaire de saisie papier, je ne peux pas le modifier.

Mon but final est que les réponses au questionnaire crée alimente le fichier excel et que le répondant puisse récupérer son rapport sous pdf ou excel.

J'ai réalisé un questionnaire Google forms avant de me rendre compte que les actions conditionelles ne me conviennent pas.

En actions conditionelles j'ai besoin:

\- que des réponses se remplissent en fonction de la réponse à une question précédente

de sauter des questions

d'ajouter des questions

Je ne m'y connais pas du tout en code, (rien de rien)

Je souhaiterais utliser en priorité SurveyJS sauf si vous me proposez une solution plus adéquate.

Merci d'avance pour votre aide

Je reste à l'écoute pour répondre à vos questions concernant mon projet.

https://redd.it/1vm9srv
@r_devops

Vendor‑Neutral or Cloud‑Specific for Kubernetes: What’s the Better Path for a New Container Learner?

I have ADHD and I wasn’t taking my medicine like I should have been, so why I started learning containers out of the blue may have something to do with that. However, I started back on my meds and I am still learning and highly interested in building and managing containers. I have only been doing this for two weeks so I’m still very new to it. I still have to look up Git commands and still run into a lot of simplistic issues that I am fixing through troubleshooting. But I have to say, building and managing containers via Docker Desktop on my personal laptop is exciting and fun. I don’t know if I’ll turn this into a career. I am currently in IT, and have been for 10 years but I am 40 years old. I’m sure my age would put a lot of people off no matter how much knowledge I have and skill I can show.

My next step is going to be learning Kubernetes. And this is where I am asking for advice on. Should I continue my learning via the vendor neutral approach or should I move my learning to a cloud platform? There seems to be some advantages to learning a specific cloud platform but if the knowledge I’m building now translates to all of them 1:1, then sticking with vendor neutral learning would be more advantageous in the long run.

Also, if you guys don’t mind answering this question as well, I’d greatly appreciate it. What else should I learn?

https://redd.it/1vmdexs
@r_devops

Cheapest server for a school project

Hello guys, i have a school project to deploy an app on cloud using ansible, the the servers are not given by the school itself.
What is the cheapest available clouds out there? I only need it to validate the project and then I can cancel.

https://redd.it/1vmgrns
@r_devops

AI Agent for CI Failure Diagnosis

I'm trying to build an agent that finds the reason why CI run failed

Like, what are the things going on in your head and

Let's imagine. Your brain has something like a belief system that sort of makes you suspect certain things more.

I want to know how practitioners think, like what doubts they are holding the moment they see a CI file, and how they determine what to look for to find the real cause and if it's a real code failure or some flaky test

And finally, At what point do they determine that they're pretty much sure about the cause?

https://redd.it/1vmg4ew
@r_devops

Deploying docker-compose.yml

Hello all. The circumstance I have working with is the following:
* I have an Apache2 PHP server that gets bundled as a Docker image in a CI process to ECR

* I have an infra repository with a docker-compose.yml that bundles the PHP Docker image to an Nginx image, alongside Nginx config like attaching TLS certs


When the CICD process deploys a release, it deploys a new EC2 with a given user data script to prop up the server. If I only had a Docker image, the user data would generally look like "Pull down ECR image and start image", however in this case I am spinning up a docker-compose.yml file.

How is this typically done? I suppose I *can* add a CI process to zip up the docker-compose.yml and related nginx config, however feels backwards? Is there a consensus with this?

If I am fundamentally misunderstanding something let me know, I'd say my only constraint is I'd like to solve this problem in a relatively cloud agnostic environment (so keeping EC2 as a VM, ECR as a registry, but excluding abstractions like Fargate or ECS)

Thanks!

https://redd.it/1vmjm45
@r_devops

Has anyone looked at RapidFort's curated libraries as a Chainguard alternative?

We've been evaluating Chainguard libraries as a way to reduce the risk of pulling vulnerable or malicious open-source packages into our builds. It seemed like one of the few established options. That was until I saw that RapidFort recently launched curated libraries as well.



Has anyone had a chance to compare the two yet? I'm interested in how they differ in practice rather than on the marketing pages. Things like package coverage, malware scanning and workflow integration. Also developer experience and ongoing maintenance all matter to us.



Has anyone here evaluated both or decided one was a better fit?

https://redd.it/1vmo871
@r_devops

Tired of juggling WindTerm / Termius / FinalShell — so I made one Rust app that aims for all three

https://github.com/vesaaa/vsterm

|Capability|VsTerm|WindTerm|Termius|FinalShell|MobaXterm|SecureCRT|Xshell|Tabby|
|:-|:-|:-|:-|:-|:-|:-|:-|:-|
|Implementation language|Rust|C/C++|Electron|Java|C++|C++|C++|Electron|
|Max terminal scrollback lines|100k / 500k (Pro)|unlimited|\-|\-|360,000|128,000|\~2.1B|25,000|
|Command-block folding / outline in terminal output|✅|✅|✗|✗|✗|✗|✗|✗|
|Integrated SFTP pane / remote file manager|✅|✅|✅|✅|✅|◐|◐|◐|
|SFTP transfer progress / queue visibility|✅|◐|◐|✅|◐|◐|◐|◐|
|ZMODEM (rz / sz) built in|✅|✅|✗|✅|◐|✅|✅|✅|
|ZMODEM progress surfaced in the app|✅|✅|✗|◐|◐|◐|◐|◐|
|Terminal ↔ file-pane path sync|✅|✗|◐|◐|✗|✗|✗|✗|
|Elevated SFTP that can follow sudo -i / su|✅|✗|✗|✗|✗|✗|✗|✗|
|Route diagram / policy-routing topology|✅|✗|✗|✗|✗|✗|✗|✗|
|Path trace with geo / ASN enrichment|✅|✗|✗|✗|✗|✗|✗|✗|
|Built-in IP quality / reputation checks|✅|✗|✗|✗|✗|✗|✗|✗|
|CPU / memory / storage graphical monitor|✅|◐|✗|✅|✗|✗|✗|✗|
|Connection / socket monitoring panel|✅|✗|✗|◐|✗|✗|✗|✗|
|Connect effects / motion polish|✅|✗|✗|✗|✗|✗|✗|✗|
|Desk pet|✅|✗|✗|✗|✗|✗|✗|✗|

https://redd.it/1vmbzrv
@r_devops

Looking for feedback on first attempt at setting up hosting and CI/CD for a fullstack .NET and React app with PostgreSQL database, hosted on Github using Github Actions.

Hi,

I'm a Senior Software Engineer with 5YOE. I'm upskilling in DevOps and project setup and have been creating this project to learn over the last few weeks. I've set up the project from scratch. Definining the project structure, creating PR pipelines for linting, formatting, typechecking, unit tests and e2e tests (tests with real database).

Over the last few weeks, I've set up the cloud infrastructure on AWS. I first just created an ec2 instance, uploaded the build files and database docker images, installed neccessary packages and ran it on the barebones ec2 to help me understand the fundamentals.

Then I went about setting up a bit more of a mature setup with RDS for the database on a private subnet, ECS Fargate to host the containerised API, using opentofu for IaaC, setting up a deploy pipeline which uploads the new container images, and runs the migrations and deploys the API service. This was quite a steep learning curve for me and I heavily relied on Gemini for a lot of this, but I have gone back over it to make sure I fully understand it. I'd really appreaciate some feedback on my setup. I'd like to know any best practices I could have followed, any big issues with the setup I've done and how I could have done better.


Also, this is just a simple project that I'm going to switch over to self hosting with Coolify on a Hertzner server soon. But I wanted to try setting up as close to a professional setup as I could, with security and CI/CD in mind.

I've done a write up of the infrastructure here
https://github.com/JackMcBride98/DotnetSpotifyPlaylistSearchTool/blob/main/infrastructure/Infrastructure.md all of the terraform files live in the /infrastructure folder.

Here is a rough overview of the architecture (I didnt draw arrows as it got quite messy)

https://preview.redd.it/e296cv2qq3jh1.png?width=732&format=png&auto=webp&s=6028ba035caf0f7292d6c92f52b1c8d37ddbd877

https://redd.it/1vn4tdm
@r_devops

Amazon Q is saving me these days otherwise I will be having a hard time using AWS services

Thank god AWS has Amazon Q otherwise I'll still be having a very hard time even finding the environment variables of a project.

https://redd.it/1vn3ypi
@r_devops

Looking for a good free resource to learn Jenkins properly

I’m new to DevOps and looking for some good free or paid resources to learn Jenkins from the ground up.
I have around 6 years of experience as a Java developer and have used Jenkins at work, but mostly from a user/developer perspective. I’ve triggered jobs and worked with existing pipelines, but I’ve never actually built a pipeline myself or understood what’s happening behind the scenes.
I’d like to learn Jenkins properly, including things like:
How Jenkins works internally
Setting up Jenkins and agents/nodes
Freestyle jobs vs Pipeline
Declarative vs Scripted Pipeline
Jenkinsfile and pipeline stages
How builds are triggered
Credentials and secrets management
Plugins and how they work
Integration with Git, Maven, Docker, etc.
CI/CD concepts and best practices
How Jenkins distributes work to agents
Debugging failed pipelines
What actually happens from a Git commit → Jenkins → build → test → deployment
I’m comfortable with Java and software development, so I’m mainly looking for something that goes beyond a basic “how to install Jenkins” tutorial.
What resources/courses did you find genuinely useful for learning Jenkins end-to-end? Free resources are preferred, but I’m also open to paid courses if they’re really worth it.
Thanks!

https://redd.it/1vna68k
@r_devops

Where should cross-system infrastructure automation stop?

I’m working through a design where each underlying system remains authoritative for its own resources, rather than putting another source of truth above everything.

The wider runtime only decides whether an operation has enough dependency state, readiness and verification to keep moving.

>The case I’m trying to pressure-test is partial execution. Say an operation spans several systems, three parts succeed and one fails. Each individual system may still be healthy, but the overall operation is incomplete.

Would you let the wider workflow block there, or does that eventually become another control layer operators have to fight with?

Interested in examples where this kind of boundary has worked badly in production.

https://redd.it/1vnaws5
@r_devops

Devops job security with AI

Recently saw posts about how AI isn't that good in distributed systems and it gave me validation on my own experience.

When the model isn't dealing with a single codebase, when it has to connect the dots between a db, redis, other cloud infra, k8s that each already have a lot of configuration, it starts to hallucinate and could pretty much put your infra at risk.

I'm not saying you can't make it work, it does work especially on a clean slate, but when there's a lot of existing context it needs a more guidance.

Using AI lets me do 10x what a DevOps engineer could do alone but I don't think the job disappears anytime soon.

https://redd.it/1vnacqq
@r_devops