Thinkube Kubernetes

Check and restart services from your agent

Ask what is running and whether it is healthy, restart a service, and turn one off and on again, from a chat with your agent

Level
beginner
Time
30 min
Risk
low
Updated
2026-10-04

Health checksServicesMCP

Overview

Basic idea

Thinkube Control knows every service on the platform: the core services, the optional components you installed, and the apps you deployed. The agent reads and changes them through the same operations the dashboard uses.

  • Health, on a schedule. Every 2 minutes Thinkube Control checks each enabled service’s health address, and each of its other addresses. The latest result is on the service’s card and in its details.

  • Restart. A rolling restart of the service’s pods: a new pod starts, then the old one stops.

  • Turn off and on. An optional component or an app can be scaled to zero and back. Core services stay on.

What you’ll accomplish

You list the services, read the health and pods of two of them, restart the docs app, and turn it off and on again.

What to know before starting

Required

  • Asking your agent in Thinkube IDE to do things on the platform.

Optional

Supported hardware

  • GPU: not needed. The operations work on any hardware.

  • Architecture: amd64 or arm64.

Prerequisites

Platform

  • Thinkube running, with Thinkube IDE open. Ask your agent: "what’s running?"

  • An app of your own to restart, for example the one from Build a web app from the template. The reference run, one run on the test cluster this page was checked on, used docs.

Components

  • Qdrant, for Step 2. Ask your agent: "is Qdrant installed?" Any installed service works the same way.

Instructions

Step 1. List the services

Ask your agent:

› what's running?

Or by hand: in Thinkube Control open All Services.

Expected output, from the reference run, shortened:

40 services
core:      Argo Workflows, ArgoCD, JuiceFS, Thinkube Notebooks, Thinkube Identity, Thinkube Packages,
           Thinkube Git, GPU Operator, Thinkube IDE, PostgreSQL, Thinkube Registry, Thinkube Storage,
           Thinkube Experiments
optional:  Chroma, Knative, ClickHouse, NATS, Ollama, Langflow, Langfuse, OpenSearch Dashboards,
           Qdrant, Prometheus, Perses, pgAdmin, Valkey
component: Text Embeddings, Vllm, Tensorrt
user_app:  Docs, Todo, Notes, Files, ... (Tasks and Knative Demo turned off)

The list says what each service is and whether it is on. Health is in the details.

Step 2. Check a service’s health

Ask your agent:

› is Qdrant healthy?

Or by hand: in All Services, open the service’s card.

Expected output, from the reference run:

Qdrant: healthy, checked 00:13:33
  api: healthy, 200 in 40 ms
pod qdrant-0 on tkamd2: Running, ready, 0 restarts
requests 2 CPU and 4Gi, limits 4 CPU and 8Gi

A service with several addresses lists each one with its own result. Read those lines as well as the overall status.

Step 3. Restart a service

Ask your agent:

› restart docs

Or by hand: on the service’s card or details page, choose Restart Service.

Expected output, from the reference run:

Service Docs restarted successfully

and in the details a moment later:

recent action: restart at 00:12:53
pod docs-docs-6545b67686-4d4pp on tkamd2: Running, ready, started 00:12:55
health: healthy, checked 00:13:33

Step 4. Turn it off and on again

Turning a service off scales it to zero and keeps everything else: its data, its address, its settings. Turning it on starts it with the replicas it had.

Ask your agent:

› turn off docs

Or by hand: on the details page, switch Service Enabled off.

Expected output, from the reference run:

is_enabled: false
recent action: disable at 00:14:00

No copy of docs was running, and https://docs.<your domain>/ answered 503.

Then:

› turn docs back on

Expected output, from the reference run:

is_enabled: true
recent action: enable at 00:14:06
pod docs-docs-6545b67686-jpdvz on tkspark: Pending

The site answered 200 again at 00:14:09.

Step 5. Next steps

To undo: Turn the service on again.

Troubleshooting

Symptom Cause Fix

Turning the service off answers 400 with Core services cannot be disabled

The service is a core service, such as JuiceFS; the platform needs it running

Restart it instead, with "restart <service>".