Stack Recipes

Introduction to Stack recipes which help you deploy a full MLOps stack in minutes!

This is an older version of the ZenML documentation. To read and view the latest version please visit this up-to-date URL.

When we first created ZenML as an extensible MLOps framework for creating portable, production-ready MLOps pipelines, we saw many of our users having to deal with the pain of deploying infrastructure from scratch to run these pipelines.

The expertise of setting up these often-complex stacks up shouldn't be a prerequisite to running your ML pipelines. We created stack recipes as a way to allow you to quickly get started with a full-fledged MLOps stack with the execution of just a couple of simple commands. Read on to learn what a recipe is, how you can deploy it and the steps needed to create your own!

Stack Recipes ๐Ÿฑ

A Stack Recipe is a collection of carefully-crafted Terraform modules and resources which, when executed, creates a range of stack components that can be used to run your pipelines. Each recipe is designed to offer a great deal of flexibility in configuring the resources while preserving the ease of application through the use of sensible defaults.

Check out the full list of available recipes at the mlops-stacks repository.

Deploying a recipe ๐Ÿš€

To use the stack recipe CLI commands, you will have to install some optional dependencies with zenml. Run pip install "zenml[stacks]" to get started!

Detailed steps are available in the README of the respective stack's recipe but here's what a simple flow could look like:

  1. ๐Ÿ“ƒ List the available recipes in the repository.

    zenml stack recipe list
  2. Pull the recipe that you wish to deploy to your local system.

    zenml stack recipe pull <STACK_RECIPE_NAME>
  3. ๐ŸŽจ Customize your deployment by editing the default values in the file. This file holds all the configurable parameters for each of the stack components.

  4. ๐Ÿ” Add your secret information like keys and passwords into the values.tfvars.json file which is not committed and only exists locally.

  5. ๐Ÿš€ Deploy the recipe with this simple command.

    zenml stack recipe deploy <STACK_RECIPE_NAME>

    If you want to allow ZenML to automatically import the created resources as a ZenML stack, pass the --import flag to the command above. By default, the imported stack will have the same name as the stack recipe and you can provide your own custom name with the --stack-name option.

  6. You'll notice that a ZenML stack configuration file gets created after the previous command executes ๐Ÿคฏ! This YAML file can be imported as a ZenML stack manually by running the following command.


Deleting resources

  1. ๐Ÿ—‘๏ธ Once you're done running your pipelines, there's only a single command you need to execute that will take care of cleaning up all the resources that you had created on your cloud.

    zenml stack recipe destroy <STACK_RECIPE_NAME>
  2. (Optional) ๐Ÿงน You can also remove all the downloaded recipe files from the pull execution by using the clean command.

    zenml stack recipe clean

Check out the API docs to learn more about each of these commands and the options that are available.

Further Integration with the ZenML CLI ๐Ÿ™

The ZenML CLI offers a set of commands to make it easy for you to list, pull and deploy recipes from anywhere!

In addition to the underlying terraform functionality, these commands also offer the following:

  • ability to list all the available recipes conveniently before you choose to deploy any one of them.

  • checks to ensure that you have all the binaries/tools installed for running a recipe.

  • extensive logs and error messages that guide you in case any of the recipes fails or misbehaves.

  • option to automatically import a ZenML stack out of the components created after deploying a stack recipe.

Creating your own recipe ๐Ÿง‘โ€๐Ÿณ

The number of recipes available right now is finite and there can be combinations of stack components that are not yet covered by any of the existing recipes. If you wish, you can contribute a recipe for any combination that you'd like to see.

The file on the repository lists the principles that each recipe follows and gives details about the steps you should take when designing your own recipe. Feel free to also reach out to the ZenML community on Slack ๐Ÿ‘‹ if you need help with any part of the process!

Manual Recipes with Terraform

You can still use the ZenML stack recipes without needing the zenml stack recipe CLI commands or even without installing ZenML. Since each recipe is a group of Terraform modules, you can simply use the Terraform CLI to perform apply and destroy operations.

Create the stack

  1. ๐ŸŽจ Customize your deployment by editing the default values in the file.

  2. ๐Ÿ” Add your secret information like keys and passwords into the values.tfvars.json file which is not committed and only exists locally.

  3. Initialize Terraform modules and download provider definitions.

    terraform init
  4. Apply the recipe.

    terraform apply

Getting the outputs

For outputs that are sensitive, you'll see that they are not shown directly on the logs. To view the full list of outputs, run the following command:

terraform output

To view individual sensitive outputs, use the following format. Here, the metadata password is being obtained.

terraform output metadata-db-password

Deleting resources (manually)

  1. ๐Ÿ—‘๏ธ Run the destroy function to clean up all resources.

    terraform destroy

Last updated