Kubernetes Shim Design

Github repo: https://github.com/apache/yunikorn-k8shim

Please read the architecture doc before reading this one, you will need to understand the 3 layer design of YuniKorn before getting to understand what is the Kubernetes shim.

The Kubernetes shim

The YuniKorn Kubernetes shim is responsible for talking to Kubernetes, it is responsible for translating the Kubernetes cluster resources, and resource requests via scheduler interface and send them to the scheduler core. And when a scheduler decision is made, it is responsible for binding the pod to the specific node. All the communication between the shim and the scheduler core is through the scheduler-interface.

The admission controller

The admission controller runs in a separate pod, it runs a mutation webhook and a validation webhook, where:

1. The mutation webhook mutates pod spec by:
   • Adding schedulerName: yunikorn
     • By explicitly specifying the scheduler name, the pod will be scheduled by YuniKorn scheduler.
   • Adding applicationId label
     • When a label applicationId exists, reuse the given applicationId.
     • When a label spark-app-selector exists, reuse the given spark app ID.
     • Otherwise, assign a generated application ID for this pod, using convention: yunikorn-<namespace>-autogen. This is unique per namespace.
   • Adding queue label
     • When a label queue exists, reuse the given queue name. Note, if placement rule is enabled, values set in the label is ignored.
     • Otherwise, adds queue: root.default
   • Adding disableStateAware label
     • If pod was assigned a generated applicationId by the admission controller, also set disableStateAware: true. This causes the generated application to immediately transition from the Starting to Running state so that it will not block other applications.
2. The validation webhook validates the configuration set in the configmap
   • This is used to prevent writing malformed configuration into the configmap.
   • The validation webhook calls scheduler validation REST API to validate configmap updates.

Admission controller deployment

By default, the admission controller is deployed as part of the YuniKorn Helm chart installation. This can be disabled if necessary (though not recommended) by setting the Helm parameter embedAdmissionController to false.

On startup, the admission controller performs a series of tasks to ensure that it is properly registered with Kubernetes:

1. Loads a Kubernetes secret called admission-controller-secrets. This secret stores a pair of CA certificates which are used to sign the TLS server certificate used by the admission controller.
2. If the secret cannot be found or either CA certificate is within 90 days of expiration, generates new certificate(s). If a certificate is expiring, a new one is generated with an expiration of 12 months in the future. If both certificates are missing or expiring, the second certificate is generated with an expiration of 6 months in the future. This ensures that both certificates do not expire at the same time, and that there is an overlap of trusted certificates.
3. If the CA certificates were created or updated, writes the secrets back to Kubernetes.
4. Generates an ephemeral TLS server certificate signed by the CA certificate with the latest expiration date.
5. Validates, and if necessary, creates or updates the Kubernetes webhook configurations named yunikorn-admission-controller-validations and yunikorn-admission-controller-mutations. If the CA certificates have changed, the webhooks will also be updated. These webhooks allow the Kubernetes API server to connect to the admission controller service to perform configmap validations and pod mutations.
6. Starts up the admission controller HTTPS server.

Additionally, the admission controller also starts a background task to wait for CA certificates to expire. Once either certificate is expiring within the next 30 days, new CA and server certificates are generated, the webhook configurations are updated, and the HTTPS server is quickly restarted. This ensures that certificates rotate properly without downtime.

In production clusters, it is recommended to deploy the admission controller with two replicas by setting the Helm parameter admissionController.replicaCount to 2. This will ensure that at least one admission controller webhook is reachable by the Kubernetes API server at all times. In this configuration, the CA certificates and webhook configurations are shared between the instances.