CalculateAvailableReplicas calculates available worker replicas at the cluster level A worker is available if its Pod is running.

CalculateDesiredReplicas calculate desired worker replicas at the cluster level.

CalculateMaxReplicas calculates max worker replicas at the cluster level.

CalculateMinReplicas calculates min worker replicas at the cluster level.

CalculatePodResource returns the total resources of a pod.

CalculateReadyReplicas calculates ready worker replicas at the cluster level A worker is ready if its Pod has a PodCondition with type == Ready and status == True.

CheckAllPodsRunning returns true if all the RayCluster's Pods are running, false otherwise.

CheckLabel makes sure the label value does not start with a punctuation and the total length is < 63 char.

CheckName makes sure the name does not start with a numeric value and the total length is < 63 char.

CompareJsonStruct This is a way to better compare if two objects are the same when they are json/yaml structs.

EnvVarByName returns an entry in []corev1.EnvVar that matches a name.

ExtractRayIPFromFQDN extracts the head service name (i.e., RAY_IP, deprecated) from a fully qualified domain name (FQDN).

FetchHeadServiceURL fetches the URL that consists of the FQDN for the RayCluster's head service and the port with the given port name (defaultPortName).

FindContainerPort searches for a specific port $portName in the container.

FormatInt returns the string representation of i in the given base, for 2 <= base <= 36.

GenerateFQDNServiceName generates a Fully Qualified Domain Name.

GenerateHeadServiceName generates a Ray head service name.

GenerateIdentifier generates identifier of same group pods.

GenerateIngressName generates an ingress name from cluster name.

Json-serializes obj and returns its hash string.

GenerateRayClusterName generates a ray cluster name from ray service name.

GenerateRayJobId generates a ray job id for submission.

GenerateRouteName generates an ingress name from cluster name.

GenerateServeServiceLabel generates label value for serve service selector.

GenerateServeServiceName generates name for serve service.

GetClusterDomainName returns cluster's domain name.

GetHeadGroupServiceAccountName returns the head group service account if it exists.

GetNamespace return namespace.

IsCreated returns true if pod has been created and is maintained by the API server.

IsJobFinished checks whether the given Job has finished execution.

IsRunningAndReady returns true if pod is in the PodRunning Phase, if it has a condition of PodReady.

PodGenerateName returns the value that should be used for a Pod's generateName based on the RayCluster name and node type (head or worker).

RayOriginatedFromCRDLabelValue generates a value for the label RayOriginatedFromCRDLabelKey This is also the only function to construct label filter of resources originated from a given CRDType.

The default application name.

The default name for kuberay operator.

Head Pod event list.

Ingress event list.

RayJob event list.

Redis Cleanup Job event list.

Role event list.

RoleBinding list.

Route event list.

Service event list.

ServiceAccount event list.

Worker Pod event list.

Use as separator for pod name, for example, raycluster-small-size-worker-0.

Use as default port.

Probe timeout for Head pod needs to be longer as it queries two endpoints (api/local_raylet_healthz & api/gcs_healthz).

Probe timeout for Head pod needs to be longer as it queries two endpoints (api/local_raylet_healthz & api/gcs_healthz).

Ray FT default liveness probe values.

Ray FT default readiness probe values.

For Ray >= 1.11.0, "DefaultRedisPort" actually refers to the GCS server port.

Default application name.

The default AppProtocol for Kubernetes service.

Ray core default configurations.

If set to true, the RayJob CR itself will be deleted if shutdownAfterJobFinishes is set to true.

Generic Pod event list.

This KubeRay operator environment variable is used to determine if the Redis cleanup Job should be enabled.

This environment variable for the KubeRay operator is used to determine whether to enable the injection of readiness and liveness probes into Ray head and worker containers.

This KubeRay operator environment variable is used to determine if random Pod deletion should be enabled.

If set to true, kuberay creates a normal ClusterIP service for a Ray Head instead of a Headless service.

EnableServeServiceKey is exclusively utilized to indicate if a RayCluster is directly used for serving.

Finalizers for GCS fault tolerance.

The default suffix for Headless Service for multi-host worker groups.

KUBERAY_VERSION is the build version of KubeRay.

Environment variables for Ray Autoscaler V2.

Use as container env variable.

Environment variables for RayJob submitter Kubernetes Job.

The value of RAY_NODE_TYPE_NAME is the name of the node group (i.e., the value of the "ray.io/group" label).

Ray health check related configurations Note: Since the Raylet process and the dashboard agent process are fate-sharing, only one of them needs to be checked.

In KubeRay, the Ray container must be the first application container in a head or worker Pod.

Ray GCS FT related annotations.

Finalizers for RayJob.

RayNodeHeadGroupLabelValue is the value for the RayNodeGroupLabelKey label on a head node.

RayOriginatedFromCRNameLabelKey and RayOriginatedFromCRDLabelKey are the labels used to associate the root KubeRay Custom Resource.

If this annotation is set to "true", the KubeRay operator will not modify the container's command.

Batch scheduling labels TODO(tgaddair): consider making these part of the CRD.

These are markers used by the calculateStatus() for setting the RayClusterReplicaFailure condition.

These are markers used by the calculateStatus() for setting the RayClusterReplicaFailure condition.

These are markers used by the calculateStatus() for setting the RayClusterReplicaFailure condition.

These are markers used by the calculateStatus() for setting the RayClusterReplicaFailure condition.

These are markers used by the calculateStatus() for setting the RayClusterReplicaFailure condition.

Job URL paths.

Multi-application URL paths.

RayJobInfo is the response of "ray job status" api.

RayJobRequest is the request body to submit.

Describes the status of an application.

V2 Serve API Response format.

ServeDeploymentStatus and ServeApplicationStatus describe the format of status(es) that will be returned by the GetMultiApplicationStatus method of the dashboard client Describes the status of a deployment.

TODO (kevin85421): Define CRDType here rather than constant.go to avoid circular dependency.

Currently, KubeRay fires events when failures occur during the creation or deletion of resources.