Big News!  Tilt is joining Docker

Modifying YAML for Dev

Kubernetes lets you define your running services with declarative YAML.

It’s really common to want to make a few tweaks to that YAML for dev, e.g.,

  • Set an env variable like DEBUG=true.

  • Set the number of servers to exactly 1 (rather than auto-scaling).

  • Set a label to distinguish dev servers from prod servers.

Fortunately, there’s a great ecosystem of YAML management tools to help you! helm, kustomize, and ytt are all great.

In this guide, we’ll show you how to patch your YAML for your dev environment – either with a few lines of Tiltfile code or with one of the tools above.

What Tilt Does Automatically

Let’s start with a basic example project:

docker_build('example-html-image', '.')

When Tilt deploys this YAML, it will automatically:

  • Apply the label tilt to every resource.

  • Ensure that all containers have pullPolicy: IfPresent or pullPolicy: Never set when using a local cluster (so that you get the locally-built image). app

  • Inject a fully-resolved image tag into the resource.

If you want to drill down more, you can use the CLI to explore what YAML Tilt is aware of and what it has applied to the cluster.

To view the YAML sets in a running dev env, use:

$ tilt get kubernetesapply
tilt-site   2022-01-18T15:20:55Z
docs-site   2022-01-18T15:20:55Z
blog-site   2022-01-18T15:20:55Z

This tells us we have 3 resources that apply YAML: tilt-site, docs-site, and blog-site.

To see the YAML in a specific resource:

$ tilt get kubernetesapply docs-site -o yaml

This will print both spec.yaml (the YAML that you originally gave Tilt) and status.resultYAML (the YAML that came back from the cluster).

How to Make Small Patches in a Tiltfile

If you’re already using a tool like Helm for managing YAML, keep using it! It’s great! Tilt is NOT trying to replace it for you!

But for one-off cases where you want to make a small patch, here are a few Tiltfile APIs that you may see:

  • read_yaml_stream reads a file from disk and decodes it to a list of objects.

  • decode_yaml_stream takes a YAML string and decodes it to a list of objects.

  • encode_yaml_stream takes a list of objects and encodes them to a string.

For example, to set the namespace on a list of objects:

objects = read_yaml_stream('deployment.yaml')
for o in objects:
  o['metadata']['namespace'] = 'my-ns'

You can see this approach in practice in the namespace extension.

How to Connect Existing YAML Tools

If you have an existing shell script that produces YAML, connecting it to Tilt is easy.


Tilt will run the script on startup, register the YAML, and deploy it.

This approach is so common that there are two built-in functions for pulling YAML from existing templating tools:

Both of these functions return YAML so that you can register them with k8s_yaml.

They’re both simply wrappers around local(), but have some nice ergonomics for passing arguments and watching dependencies.

If you’d like to add new template tools, you don’t need to make them built-in functions! These tools can be packaged up as Tilt Extensions and shared with the community of Tilt users.

When YAML is Not Enough

Some modern Kubernetes tools don’t produce a pile of YAML for you to deploy yourself.

They deploy resources in a particular order, wait on health checks, or add post-deploy hooks. Then they output the YAML at the end!

For example, Pulumi lets you define your infra in Javascript or Python. Helm can download packages from a remote repository you found on Artifact Hub. And Helm packages can have chart hooks that modify the release life-cycle.

For these tools, we recommend using k8s_custom_deploy to deploy instead of k8s_yaml. For an example, see the helm_resource extension.

Further Reading