Unclear Kubernetes connection documentation

[nevarsin]

Hi,

I'm trying to follow this article https://hoop.dev/docs/quickstarts/kubernetes to try a first connection from my hoop docker-compose test env to a sample k8s cluster.

What I have trouble going trough is:

• hoop CLI doesn't look like coming with an "exec" option anymore
• Adding a connection with hoop admin create conn k8s -a system -e b64-filesystem:KUBECONFIG=$(cat $HOME/.kube/config |base64) -- kubectl doesn't look like it's adding an env var to the agent
• am I supposed to add the kubectl binary on my own to the agent?
• is running an Agent on each cluster the preferred way atm?

Thanks again!

[sandromello]

Hello @nevarsin,

Thank you for bringing this to our attention. I wanted to address your questions about our Kubernetes documentation and provide some important updates:

We are currently revising our Kubernetes quick start guide to reflect recent changes. Here are the key points you should be aware of:

1. The exec command has been deprecated. Instead, you can establish connections using either:

   • The hoop connect command
   • Our web application interface

2. Environment variables are dynamically injected at runtime and become visible only during session creation. This means you won't see them in the configuration beforehand, but they will be available once your session starts.

3. Regarding kubectl availability:

   • The system/default agent includes kubectl pre-installed
   • Any custom agents using the hoophq/hoopdev image also come with kubectl configured

4. For service exposure, you have flexibility in your deployment. You can run either single or multiple agents within the same cluster, depending on your specific service architecture requirements.

I will notify you once we have completed the documentation updates. In the meantime, please let me know if you have any questions about these changes.

[nevarsin]

Hi, thanks for your prompt answer, I really appreciate the short feedback loop from you guys, it's something nobody should take for granted.

Re the issue: I will try again now but one thing that threw me off the rails was that the https://hoop.dev/docker-compose.yml file is using the hoophq/hooplabs image for the agents while I understand that the correct one should be the one you mentioned (hooplabs/hoopdev) where, infact, I could find the kubectl binary pre-installed.

[nevarsin]

Ok I confirm I was to access a cluster by:

1. using the hoophq/hoopdev image for agents, and manually populating the HOOP_KEY env var for that container
2. I added the cluster by doing `hoop admin create conn k8s -a newagent -e b64-filesystem:KUBECONFIG=$(cat $HOME/.kube/config |base64) -- kubectl' however this led to a malformed KUBECONFIG content being parsed and stored, I opened the UI and manually pasted my kubeconfig file content into the key and manually set the kubectl command as default

• After that, a command like hoop connect k8s get pods works as expected

Thank you, I look forward to updated doc.