The Kubeflow orchestrator is an orchestrator flavor provided by the ZenML kubeflow integration that uses Kubeflow Pipelines to run your pipelines.
This component is only meant to be used within the context of a remote ZenML deployment scenario. Usage with a local ZenML deployment may lead to unexpected behavior!
When to use it
You should use the Kubeflow orchestrator if:
you're looking for a proven production-grade orchestrator.
you're looking for a UI in which you can track your pipeline runs.
you're already using Kubernetes or are not afraid of setting up and maintaining a Kubernetes cluster.
you're willing to deploy and maintain Kubeflow Pipelines on your cluster.
How to deploy it
To run ZenML pipelines on Kubeflow, you'll need to set up a Kubernetes cluster and deploy Kubeflow Pipelines on it. This can be done in a variety of ways, depending on whether you want to use a cloud provider or your own infrastructure:
Since Kubernetes v1.19, AKS has shifted to containerd. However, the workflow controller installed with the Kubeflow installation has Docker set as the default runtime. In order to make your pipelines work, you have to change the value to one of the options listed here, preferably k8sapi.
This change has to be made by editing the containerRuntimeExecutor property of the ConfigMap corresponding to the workflow controller. Run the following commands to first know what config map to change and then to edit it to reflect your new value:
kubectl get configmap -n kubeflow
kubectl edit configmap CONFIGMAP_NAME -n kubeflow
# This opens up an editor that can be used to make the change.
Have an existing Kubernetes cluster set up.
Download and installkubectl and configure it to talk to your Kubernetes cluster.
( optional) set up a Kubernetes Service Connector to grant ZenML Stack Components easy and secure access to the remote Kubernetes cluster. This is especially useful if your Kubernetes cluster is remotely accessible, as this enables other ZenML users to use it to run pipelines without needing to configure and set up kubectl on their local machines.
If one or more of the deployments are not in the Running state, try increasing the number of nodes in your cluster.
If you're installing Kubeflow Pipelines manually, make sure the Kubernetes service is called exactly ml-pipeline. This is a requirement for ZenML to connect to your Kubeflow Pipelines deployment.
How to use it
To use the Kubeflow orchestrator, we need:
A Kubernetes cluster with Kubeflow pipelines installed. See the deployment section for more information.
A ZenML server deployed remotely where it can be accessed from the Kubernetes cluster. See the deployment guide for more information.
The ZenML kubeflow integration installed. If you haven't done so, run
zenmlintegrationinstallkubeflow
Docker installed and running (unless you are using a remote Image Builder in your ZenML stack).
If you are using a single-tenant Kubeflow installed in a Kubernetes cluster managed by a cloud provider like AWS, GCP or Azure, it is recommended that you set up a Service Connector and use it to connect ZenML Stack Components to the remote Kubernetes cluster. This guarantees that your Stack is fully portable on other environments and your pipelines are fully reproducible.
The name of your Kubernetes context which points to your remote cluster. Run kubectl config get-contexts to see a list of available contexts. NOTE: this is no longer required if you are using a Service Connector to connect your Kubeflow Orchestrator Stack Component to the remote Kubernetes cluster.
We can then register the orchestrator and use it in our active stack. This can be done in two ways:
If you have a Service Connector configured to access the remote Kubernetes cluster, you no longer need to set the kubernetes_context attribute to a local kubectl context. In fact, you don't need the local Kubernetes CLI at all. You can connect the stack component to the Service Connector instead:
# List all available Kubernetes clusters that can be accessed by service connectorszenmlservice-connectorlist-resources--resource-typekubernetes-cluster-e# Register the Kubeflow orchestrator and connect it to the remote Kubernetes cluster zenml orchestrator register <ORCHESTRATOR_NAME> --flavor kubeflow --connector <SERVICE_CONNECTOR_NAME> --resource-id <KUBERNETES_CLUSTER_NAME>
# Register a new stack with the orchestratorzenml stack register <STACK_NAME> -o <ORCHESTRATOR_NAME> -a <ARTIFACT_STORE_NAME> -c <CONTAINER_REGISTRY_NAME> ... # Add other stack components as needed
The following example demonstrates how to register the orchestrator and connect it to a remote Kubernetes cluster using a Service Connector:
if you don't have a Service Connector on hand and you don't want to register one, the local Kubernetes kubectl client needs to be configured with a configuration context pointing to the remote cluster. The kubernetes_context must also be configured with the value of that context:
zenmlorchestratorregister<ORCHESTRATOR_NAME> \--flavor=kubeflow \--kubernetes_context=<KUBERNETES_CONTEXT># Register a new stack with the orchestratorzenml stack register <STACK_NAME> -o <ORCHESTRATOR_NAME> -a <ARTIFACT_STORE_NAME> -c <CONTAINER_REGISTRY_NAME> ... # Add other stack components as needed
ZenML will build a Docker image called <CONTAINER_REGISTRY_URI>/zenml:<PIPELINE_NAME> which includes all required software dependencies and use it to run your pipeline steps in Kubeflow. Check out this page if you want to learn more about how ZenML builds these images and how you can customize them.
You can now run any ZenML pipeline using the Kubeflow orchestrator:
pythonfile_that_runs_a_zenml_pipeline.py
Kubeflow UI
Kubeflow comes with its own UI that you can use to find further details about your pipeline runs, such as the logs of your steps. For any runs executed on Kubeflow, you can get the URL to the Kubeflow UI in Python using the following code snippet:
from zenml.client import Clientpipeline_run =Client().get_pipeline_run("<PIPELINE_RUN_NAME>")orchestrator_url = pipeline_run.run_metadata["orchestrator_url"].value
Additional configuration
For additional configuration of the Kubeflow orchestrator, you can pass KubeflowOrchestratorSettings which allows you to configure (among others) the following attributes:
client_args: Arguments to pass when initializing the KFP client.
user_namespace: The user namespace to use when creating experiments and runs.
pod_settings: Node selectors, affinity, and tolerations to apply to the Kubernetes Pods running your pipeline. These can be either specified using the Kubernetes model objects or as dictionaries.
Check out the SDK docs for a full list of available attributes and this docs page for more information on how to specify settings.
Enabling CUDA for GPU-backed hardware
Note that if you wish to use this orchestrator to run steps on a GPU, you will need to follow the instructions on this page to ensure that it works. It requires adding some extra settings customization and is essential to enable CUDA for the GPU to give its full acceleration.
Important Note for Multi-Tenancy Deployments
Kubeflow has a notion of multi-tenancy built into its deployment. Kubeflow's multi-user isolation simplifies user operations because each user only views and edited the Kubeflow components and model artifacts defined in their configuration.
Using the ZenML Kubeflow orchestrator on a multi-tenant deployment without any settings will result in the following error:
HTTP response body: {"error":"Invalid input error: Invalid resource references for experiment. ListExperiment requires filtering by namespace.","code":3,"message":"Invalid input error: Invalid resource references for experiment. ListExperiment requires filtering by
namespace.","details":[{"@type":"type.googleapis.com/api.Error","error_message":"Invalid resource references for experiment. ListExperiment requires filtering by namespace.","error_details":"Invalid input error: Invalid resource references for experiment. ListExperiment requires filtering by namespace."}]}
In order to get it to work, we need to leverage the KubeflowOrchestratorSettings referenced above. By setting the namespace option, and by passing in the right authentication credentials to the Kubeflow Pipelines Client, we can make it work.
First, when registering your Kubeflow orchestrator, please make sure to include the kubeflow_hostname parameter. The kubeflow_hostnamemust end with the /pipeline post-fix.
zenmlorchestratorregister<NAME> \--flavor=kubeflow \--kubeflow_hostname=<KUBEFLOW_HOSTNAME># e.g. https://mykubeflow.example.com/pipeline
Then, ensure that you use the pass the right settings before triggering a pipeline run. The following snippet will prove useful:
import requestsfrom zenml.client import Clientfrom zenml.integrations.kubeflow.flavors.kubeflow_orchestrator_flavor import ( KubeflowOrchestratorSettings,)NAMESPACE ="namespace_name"# This is the user namespace for the profile you want to useUSERNAME ="admin"# This is the username for the profile you want to usePASSWORD ="abc123"# This is the password for the profile you want to use# Use client_username and client_password and ZenML will automatically fetch a session cookiekubeflow_settings =KubeflowOrchestratorSettings( client_username=USERNAME, client_password=PASSWORD, user_namespace=NAMESPACE)# You can also pass the cookie in `client_args` directly# kubeflow_settings = KubeflowOrchestratorSettings(# client_args={"cookies": session_cookie}, user_namespace=NAMESPACE# )@pipeline( settings={"orchestrator": kubeflow_settings }):...if"__name__"=="__main__":# Run the pipeline
Note that the above is also currently not tested on all Kubeflow versions, so there might be further bugs with older Kubeflow versions. In this case, please reach out to us on Slack.
Using secrets in settings
The above example encoded the username and password in plain text as settings. You can also set them as secrets.
# Use client_username and client_password and ZenML will automatically fetch a session cookiekubeflow_settings =KubeflowOrchestratorSettings( client_username="{{kubeflow_secret.username}}", # secret reference client_password="{{kubeflow_secret.password}}", # secret reference user_namespace="namespace_name")
See full documentation of using ZenML secrets here.
For more information and a full list of configurable attributes of the Kubeflow orchestrator, check out the SDK Docs .