Tips & Tricks for the week

I recently encountered a production bug that I couldn't reproduce locally, which was complicated by internal microservices, running memory-intensive machine learning models on GPUs, that weren't exposed through the public API.

To solve this, I used Codezero, where I could access the kubernetes cluster running the microservice from my local machine using just a DNS name.

Installed the Codezero agent using brew.

brew install c6o/tap/codezero-app

Logged into my Codezero account, created a new teamspace and installed it into my kubernetes cluster.

helm repo add --force-update codezero https://charts.codezero.io && \
helm install --create-namespace --namespace=codezero \
--set space.name='testone' --set org.id='my org id' \
--set org.apikey='this is a secret' \
--set hub.url='https://hub.codezero.io' codezero codezero/codezero

From the Codezero app, I can consume the service I need by DNS name and curl returns what I need.

curl http://nval-ltf-model-web.nval-app:8000/

{
  "collections": "http://nval-ltf-model-web.nval-app:8000/collections/",
  "greenlists": "http://nval-ltf-model-web.nval-app:8000/greenlists/"
}

For more, see my blog post: Debugging with Codezero​

What this week has in store

Adding AI tools to my chatbot has been great; I can give it functions to perform specific tasks, such as querying my database. This week, I'm going to integrate MCP servers into the mix and see how they work in conjunction with the existing tools.

See you next week!

Mark C Allen

1:1 Chat Need a Different Perspective Do you have a problem with your release process? Has Kubernetes got you down? Do you need an outsider's perspective on what's holding up your deployments? If you have 25 minutes, book a time with me. Book some time