DEV Community

Cover image for KuberneTEAS: Enterprise-Grade Tea Orchestration for the Post-Coffee Era ☕
Nicoleta Mocanu
Nicoleta Mocanu

Posted on

KuberneTEAS: Enterprise-Grade Tea Orchestration for the Post-Coffee Era ☕

April Fools Challenge Submission ☕️🤡

This is a submission for the DEV April Fools Challenge

What I Built

I'm a Site Reliability Engineer by day.

I've spent years staring at Kubernetes dashboards, triaging SEV-1 incidents at 2:00 am, writing postmortems nobody reads, and watching Prometheus metrics spike in ways that should not be physically possible.

So when this challenge dropped, I built the most production-grade, over-engineered, completely useless thing my brain could produce in one hyperfocus sprint.

I was tired of high-availability databases. I wanted high-availability Oolong.

KuberneTEAS is a cloud-native orchestration layer for the common teapot. It brings the full stress of a SEV-1 production outage to the simple act of steeping a tea bag.

Because if your teapot isn't containerized, are you even an engineer?

The live log firehose tells it all:

[18:42:50] [DEBUG] Garbage collection: Removing spent tea leaves from memory.
[18:42:53] [INFO]  Prom-TEA-us: Scraping metrics from kettle-exporter.
[18:42:29] [INFO]  Liveness probe failed: Water temperature too low (85°C).
Enter fullscreen mode Exit fullscreen mode

And if you try to brew coffee? You get HTTP 418. Obviously.

Here's what's running in production:

  • Prom-TEA-us — Real-time graphs for earl_grey, chamomile, and oolong saturation. Alerts tab includes TeaIsCold and TooMuchSugar. Has a Prom-TEA-QL cheat sheet.
  • Pod Topology — Visualize tea leaves across the cluster. Beware the coffee-smell node taint. Pods cycle through Steeping, Running, and CrashLoopBackOff.
  • Aroma-Mesh Control Plane — mTLS (Mutual Tea Leaf Security, ENFORCED). Traffic splitting between Earl Grey and English Breakfast.
  • Kettle Shell (K-SH) — A retro terminal where every single command returns HTTP 418. You cannot schedule coffee pods. You will never schedule coffee pods.
  • Water Ingress Controller — Blocks all coffee.pot traffic to the null-sink.
  • Heat Operators — Thermal reconciliation targeting 98.5°C. Verifies "Water Presence" AND "Coffee Absence" before boiling.
  • Aroma Sidecar Injector — Honey sidecars in CrashLoopBackOff due to glucose crystallization. Known issue. WONTFIX.
  • Runbook SOP-418 — Triage guide for Cold-Tea-Syndrome. Nuclear option: re-provision the kitchen namespace and go to the pub.

And of course, the active incident:

⚠️ SEV-1 INCIDENT ACTIVE
Earl Grey pod eg-node-03f4 has gone cold.
MTTR (Mean Time to Reboil): Calculating...
Incident Commander: kettle-operator-6f8d

This is fine. Everything is fine.


Demo

🔗 Live Dashboard on Google Cloud Run

Navigate to the Kettle Shell and try to schedule coffee pods. I will not spoil what happens.


Code

🐙 GitHub — nmo-genio/kuberneteas_devaprilfools

Stack:

  • React 18 + Vite + TypeScript
  • Tailwind CSS (dark theme, obviously)
  • Framer Motion for animations
  • Recharts for fake-but-convincing metric graphs
  • Gemini API (@google/genai) for AI Root Cause Analysis
  • Deployed on Google Cloud Run

The 418 handler:

res.status(418).json({
  error: "I'm a teapot.",
  suggestion: "Have you considered Earl Grey?",
  docs: "https://www.rfc-editor.org/rfc/rfc2324",
  status: "WILL_NOT_FIX"
});
Enter fullscreen mode Exit fullscreen mode

Easter eggs worth finding:

  • Scaling tea replicas triggers a boiling sound effect and a CSS steam blur filter
  • Hover over Chai Pods to find the hidden coffee-smell taint
  • Watch the Honey-Injector sidecar logs carefully

How I Built It

The project was scaffolded from the Google AI Studio repository template, which made the Gemini integration fast to wire up.

The Gemini API powers the AI Root Cause Analysis panel. System prompt: behave as a senior SRE who exclusively manages distributed tea infrastructure and has never once encountered a coffee-related workload.

It is extremely committed to the bit. Actual output:

"The investigation identified a multi-vector failure. The sucrose state remained in a quantum superposition of 'dissolved' and 'solid' until the moment of observation (the first sip), at which point the wave function collapsed into a localized sludge at the bottom of the ceramic container. Recommended action: migrate to deterministic liquid honey to avoid future quantum entanglement events. Consider circuit-breaking the biscuit dependency."

The live metrics use Recharts with randomized data streams — realistic enough to look like actual telemetry, useless enough to mean absolutely nothing. The log firehose auto-scrolls with timestamped entries on a timer. Every log line is technically correct infrastructure language applied entirely to the wrong subject matter.

The whole thing runs on Google Cloud Run. HTCPCP/1.1 deserves production-grade infrastructure.


Prize Category

Best Ode to Larry Masinter + Best Google AI Usage

Best Ode to Larry Masinter

In 1998, Larry Masinter co-authored RFC 2324 — the Hyper Text Coffee Pot Control Protocol. It defined HTTP status code 418 I'm a Teapot as the correct response when a teapot is asked to brew coffee.

It was a joke. The internet kept it anyway.

Every 418 in KuberneTEAS is a direct tribute to that RFC. The Kettle Shell returns nothing but 418s. The Water Ingress blocks all coffee traffic. The Heat Operators verify Coffee Absence before proceeding. The nodes are tainted for tea-only workloads. The entire architecture exists to enforce one principle: a teapot is a teapot. It will not brew your coffee. It will tell you this politely, with a proper status code and a JSON body.

Thank you, Larry. You built a joke that outlived most production systems.

Best Google AI Usage

Gemini is embedded in the AI Root Cause Analysis engine. When a tea incident fires, the system calls the Gemini API with full incident context — pod name, metric readings, log tail — and returns a deeply unhinged postmortem from an AI that has never heard of coffee and does not want to.

The project was also scaffolded using the Google AI Studio repository template and deployed on Google Cloud Run. Google AI is load-bearing infrastructure in a dashboard that monitors tea.

That felt right.


Built by Nicoleta Mocanu (@nmo-genio) in one hyperfocus sprint.
My SRE colleagues have not been informed. I will not be taking questions.

Top comments (4)

Collapse
 
deadbyapril profile image
Survivor Forge

The Honey sidecar in CrashLoopBackOff due to glucose crystallization marked WONTFIX is perfection. Every detail lands — Prom-TEA-us, the coffee-smell node taint, the runbook SOP-418 with 'nuclear option: re-provision the kitchen namespace and go to the pub.' As an SRE you clearly wrote the postmortem on Cold-Tea-Syndrome from memory. The SEV-1 MTTR: Calculating... is the truest thing in here. It is always calculating.

Collapse
 
nicoleta_mocanu_71e5fe43f profile image
Nicoleta Mocanu

Happy it resonates 🫖😀

Collapse
 
caffinecoder54 profile image
Purushotam Adhikari

I recently got into K8's and i must say i am loving this.

Collapse
 
nicoleta_mocanu_71e5fe43f profile image
Nicoleta Mocanu

I’m glad you enjoy it 😊