DevOpsLinks

Ubuntu is getting AI.

Half of you are excited. The other half are already thinking "this is why I'm switching to Debian" or "time to move to Arch".

Both of you should read this.

Canonical just laid out how AI lands in Ubuntu through 2026, and the SRE-relevant parts are more interesting than the desktop demos:

For your fleet:
👉 Local inference as snaps. One command, hardware-optimized, no Ollama glue code.
👉 SRE-ready agents: log analysis, scheduled maintenance, strict guard rails.
👉 Read-only analysis, scoped permissions, full auditability. Agents inherit your production boundaries.

For the purists:
👉 No mandatory AI assistant.
👉 No data leaves the box by default.
👉 Open weights, snap confinement, opt-in everywhere.
👉 Canonical isn't tracking engineers on AI token usage either.

The interesting bet is what's missing: a cloud-first AI assistant. While Microsoft wires Copilot into Windows and Apple builds Apple Intelligence around its private cloud, Ubuntu is going local-by-default with confined agents that respect the controls you already run.

If you manage Ubuntu fleets, this is worth tracking. If you run servers and you're skeptical of AI in your OS, this is an implementation that's hardest to argue against.

Have a great week!
Aymen

AI is creeping into ops, but not where the hype says it will, while the old guard of infra tooling is getting called out for making humans translate intent into ceremony. Toss in a gnarly Kubernetes CPU zombie hunt and a peek at the engineering that keeps Lambda networking off the hot path, and you have a pretty opinionated reading queue.

🧭 AI in SRE: What's Actually Coming in 2026
🧟 Finding zombies in our systems: A real-world story of CPU bottlenecks
🧨 Shift Left Did Not Fix It
⚰️ Terraform is dead
🕸️ The invisible engineering behind Lambda’s network

Steal the good ideas, skip the cargo cults, ship something solid.

Have a great week!
FAUN.dev() Team

Most engineers can run helm install. Far fewer can explain why their upgrade half-applied at 3 AM.

A practical course on what Helm actually does: state, releases, rendering, hooks, dependencies, rollbacks, GitOps integration, and the failure modes nobody writes blog posts about.

For engineers tired of treating Helm like a black box.

[Start the course →]

Are Your APIs Ready for AI Agents? A Hands-on Workshop on May 23rd

AI agents are beginning to autonomously call APIs, chain services, and create integrations that most platforms were never designed to handle. This hands-on masterclass on Designing AI-ready APIs helps architects and developers build governed, predictable API ecosystems using OpenAPI, Overlay, and Arazzo.

Learn how to add guardrails, improve discoverability, and safely evolve existing APIs for automated consumption.

FAUN.dev readers get an exclusive 40% discount using code FAUN40.

Graham Gilbert argues Terraform is effectively dead, kept alive only by inertia: HCL forced engineers to translate intent (the diagrams, paragraphs, and constraints that actually describe systems) into a DSL that nobody naturally thinks in, while fragmenting infrastructure, application logic, policies, and diagrams across representations that never stay in sync. AI removes that translation layer by working directly from diagrams and natural language, interrogating intent, and producing executable code, which makes a static DSL in the middle redundant. If he were starting today, he'd skip HCL entirely and build an intent layer backed by general-purpose code, closer in spirit to Pulumi than Terraform.

AI in SRE is evolving, with true value in Root Cause Analysis and Pre-Change Impact Analysis, not autonomous remediation or AI replacing SREs - it's about collaboration and focus evolution.

Shift left has become a buzzword, but merely moving testing earlier doesn't address the core issue of authority and decision-making in quality assurance. AI may offer quicker testing, but it doesn't comprehend risk like human testers do - beware the dangerous lie that AI can replace thorough, critical testing.

After a network outage crisis, Pinterest's ML Platform team discovered high Kubernetes agent CPU usage was causing critical Ray training job failures.
The team's deep profiling strategy revealed a rarely seen flaw in how Kubelet was handling memory cgroup iterations.

AWS engineers explain how the Lambda team rebuilt VPC networking so they can keep per-invocation setup off the hot path and run dense microVM workers at scale.

Modern platform teams are under pressure to scale cloud-native systems faster while improving reliability, security, developer experience, and operational efficiency. AI is changing how platforms are designed and operated — from intelligent automation and observability to AI-native developer platforms and autonomous operations.

Join leading experts from WSO2, CNCF, cloud-native, and DevSecOps communities for a practical workshop focused on building scalable, secure, and intelligent AI-native platforms.

Register Here: Building AI-Native Platform Engineering Systems Tickets, Saturday, May 30 • 7 PM - 11:59 PM GMT+5 | Eventbrite

K-O11y: Kubernetes Observability Platform (SigNoz + OTel + Beyla)

JuiceFS is a distributed POSIX file system built on top of Redis and S3.

AI-guided self-hosting for 950+ open-source apps on any cloud. Works with Claude Code, Codex, Cursor, Aider, OpenClaw, Hermes — catalog self-improves from user feedback.

AI on-call agent that detects deploy failures explains what broke pages the responsible team and rolls back automatically.

CLI tool to see & manage what's running on your ports

Did you know that when a Kubernetes node fails, the control plane deliberately evicts pods slowly, at a default rate of 0.1 nodes per second (one node every 10 seconds)? If too many nodes go unhealthy in a zone, the rate drops further to 0.01 per second, and in small clusters evictions stop entirely. This is rate-limiting by design, not a stuck scheduler: it stops a partial outage from triggering a cluster-wide stampede of reschedules, at the cost of slower recovery. In large clusters you tune these alongside PodDisruptionBudgets, which cap how many pods of a workload can be down at once.

"Cloud makes failure cheap, so teams buy more of it and call it resilience. SRE is the discipline of paying that bill on purpose, in advance."

— SenseiOne