Deploy with ZenML CLI
Deploying ZenML on cloud using the ZenML CLI.
The easiest and fastest way to get running on the cloud is by using the deploy
CLI command. It currently only supports deploying to Kubernetes on managed cloud services. You can check the overview page to learn about other options that you have.
Before we begin, it will help to understand the architecture around the ZenML server and the database that it uses. Now, depending on your setup, you may find one of the following scenarios relevant.
Option 1: Starting from scratch
If you don't have an existing Kubernetes cluster, you have the following two options to set it up:
Using a stack recipe that sets up a cluster along with other tools that you might need in your cloud stack like artifact stores and secret managers. Take a look at all available stack recipes to see if there's something that works for you.
Once you have created your cluster, make sure that you configure your kubectl client to talk to it.
You're now ready to deploy ZenML! Run the following command:
You will be prompted to provide a name for your deployment and details like what cloud provider you want to deploy to — and that's it! It creates the database and any VPCs, permissions, and more that are needed.
In order to be able to run the deploy
command, you should have your cloud provider's CLI configured locally with permissions to create resources like MySQL databases and networks.
Reasonable defaults are in place for you already and if you wish to configure more settings, take a look at the next scenario that uses a config file.
Option 2: Using existing cloud resources
Existing Kubernetes cluster
If you already have an existing cluster without an ingress controller, you can jump straight to the deploy
command above to get going with the defaults. Please make sure that you have your local kubectl
configured to talk to your cluster.
Having an existing NGINX Ingress Controller
The deploy
command, by default, tries to create an NGINX ingress controller on your cluster. If you already have an existing controller, you can tell ZenML to not re-deploy it through the use of a config file. This file can be found in the Configuration File Templates towards the end of this guide. It offers a host of configuration options that you can leverage for advanced use cases.
Check if an ingress controller is running on your cluster by running the following command. You should see an entry in the output with the hostname populated.
Set
create_ingress_controller
tofalse
.Supply your controller's hostname to the
ingress_controller_hostname
variable.Note: The address should not have a trailing
/
.You can now run the
deploy
command and pass the config file above, to it.Note: To be able to run the deploy command, you should have your cloud provider's CLI configured locally with permissions to create resources like MySQL databases and networks.
Existing hosted SQL database
If you also already have a database that you would want to use with the deployment, you can choose to configure it with the use of the config file. Here, we will demonstrate setting the database.
Fill the fields below from the config file with values from your database.
Run the
deploy
command and pass the config file above to it.Note To be able to run the deploy command, you should have your cloud provider's CLI configured locally with permissions to create resources like MySQL databases and networks.
Configuration file templates
Base configuration file
Below is the general structure of a config file. Use this as a base and then add any cloud-specific parameters from the sections below.
Feel free to include only those variables that you want to customize, in your file. For all other variables, the default values (shown above) will be used.
Cloud-specific settings
The database_username
and database_password
from the general config is used to set those variables for the AWS RDS instance.
Connecting to deployed ZenML
Immediately after deployment, the ZenML server needs to be activated before it can be used. The activation process includes creating an initial admin user account and configuring some server settings. You can do this only by visiting the ZenML server URL in your browser and following the on-screen instructions. Connecting your local ZenML client to the server is not possible until the server is properly initialized.
Once ZenML is deployed, one or multiple users can connect to it with the zenml connect
command.
If no arguments are supplied, ZenML will attempt to connect to the last ZenML server deployed from the local host using the zenml deploy
command:
In order to connect to a specific ZenML server, you can either pass the configuration as command line arguments or as a YAML file:
or
The YAML file should have the following structure when connecting to a ZenML server:
Here is an example of a ZenML server YAML configuration file:
To disconnect from the current ZenML server and revert to using the local default database, use the following command:
How does it work?
Here's an architecture diagram that shows how the workflow looks like when you do zenml deploy
.
The deploy CLI makes use of a "recipe" inside the zenml-io/zenml
repository to deploy the server on the right cloud. Any configuration that you pass with the CLI, is sent to the recipe as input variables.
Last updated