Files

188 lines
8.7 KiB
Go

package utils
const (
// Default application name
DefaultServeAppName = "default"
// Belows used as label key
// RayOriginatedFromCRNameLabelKey and RayOriginatedFromCRDLabelKey are the labels used to associate the root KubeRay Custom Resource.
// [Example 1] If we create a RayJob named `myjob`, then (1) the RayCluster and (2) the submitter K8s Job will have a
// `ray.io/originated-from-cr-name=myjob` and a `ray.io/originated-from-crd=RayJob` label.
//
// [Example 2] If we create a RayService named `mysvc`, then (1) the RayCluster and (2) the Kubernetes services managed by the RayService
// will have a `ray.io/originated-from-cr-name=mysvc` and a `ray.io/originated-from-crd=RayService` label.
RayOriginatedFromCRNameLabelKey = "ray.io/originated-from-cr-name"
RayOriginatedFromCRDLabelKey = "ray.io/originated-from-crd"
RayClusterLabelKey = "ray.io/cluster"
RayNodeTypeLabelKey = "ray.io/node-type"
RayNodeGroupLabelKey = "ray.io/group"
RayNodeLabelKey = "ray.io/is-ray-node"
RayIDLabelKey = "ray.io/identifier"
RayClusterServingServiceLabelKey = "ray.io/serve"
RayClusterHeadlessServiceLabelKey = "ray.io/headless-worker-svc"
HashWithoutReplicasAndWorkersToDeleteKey = "ray.io/hash-without-replicas-and-workers-to-delete"
NumWorkerGroupsKey = "ray.io/num-worker-groups"
// In KubeRay, the Ray container must be the first application container in a head or worker Pod.
RayContainerIndex = 0
// Batch scheduling labels
// TODO(tgaddair): consider making these part of the CRD
RaySchedulerName = "ray.io/scheduler-name"
RayPriorityClassName = "ray.io/priority-class-name"
// Ray GCS FT related annotations
RayFTEnabledAnnotationKey = "ray.io/ft-enabled"
RayExternalStorageNSAnnotationKey = "ray.io/external-storage-namespace"
// If this annotation is set to "true", the KubeRay operator will not modify the container's command.
// However, the generated `ray start` command will still be stored in the container's environment variable
// `KUBERAY_GEN_RAY_START_CMD`.
RayOverwriteContainerCmdAnnotationKey = "ray.io/overwrite-container-cmd"
// Finalizers for GCS fault tolerance
GCSFaultToleranceRedisCleanupFinalizer = "ray.io/gcs-ft-redis-cleanup-finalizer"
// EnableServeServiceKey is exclusively utilized to indicate if a RayCluster is directly used for serving.
// See https://github.com/ray-project/kuberay/pull/1672 for more details.
EnableServeServiceKey = "ray.io/enable-serve-service"
EnableServeServiceTrue = "true"
EnableRayClusterServingServiceTrue = "true"
EnableRayClusterServingServiceFalse = "false"
KubernetesApplicationNameLabelKey = "app.kubernetes.io/name"
KubernetesCreatedByLabelKey = "app.kubernetes.io/created-by"
// Use as separator for pod name, for example, raycluster-small-size-worker-0
DashSymbol = "-"
// Use as default port
DefaultClientPort = 10001
// For Ray >= 1.11.0, "DefaultRedisPort" actually refers to the GCS server port.
// However, the role of this port is unchanged in Ray APIs like ray.init and ray start.
// This is the port used by Ray workers and drivers inside the Ray cluster to connect to the Ray head.
DefaultRedisPort = 6379
DefaultDashboardPort = 8265
DefaultMetricsPort = 8080
DefaultDashboardAgentListenPort = 52365
DefaultServingPort = 8000
ClientPortName = "client"
RedisPortName = "redis"
DashboardPortName = "dashboard"
MetricsPortName = "metrics"
ServingPortName = "serve"
// The default AppProtocol for Kubernetes service
DefaultServiceAppProtocol = "tcp"
// The default application name
ApplicationName = "kuberay"
// The default name for kuberay operator
ComponentName = "kuberay-operator"
// The default suffix for Headless Service for multi-host worker groups.
// The full name will be of the form "${RayCluster_Name}-headless-worker-svc".
HeadlessServiceSuffix = "headless-worker-svc"
// Use as container env variable
RAY_CLUSTER_NAME = "RAY_CLUSTER_NAME"
RAY_IP = "RAY_IP"
FQ_RAY_IP = "FQ_RAY_IP"
RAY_PORT = "RAY_PORT"
RAY_ADDRESS = "RAY_ADDRESS"
REDIS_PASSWORD = "REDIS_PASSWORD"
RAY_DASHBOARD_ENABLE_K8S_DISK_USAGE = "RAY_DASHBOARD_ENABLE_K8S_DISK_USAGE"
RAY_EXTERNAL_STORAGE_NS = "RAY_external_storage_namespace"
RAY_GCS_RPC_SERVER_RECONNECT_TIMEOUT_S = "RAY_gcs_rpc_server_reconnect_timeout_s"
RAY_TIMEOUT_MS_TASK_WAIT_FOR_DEATH_INFO = "RAY_timeout_ms_task_wait_for_death_info"
RAY_GCS_SERVER_REQUEST_TIMEOUT_SECONDS = "RAY_gcs_server_request_timeout_seconds"
RAY_SERVE_KV_TIMEOUT_S = "RAY_SERVE_KV_TIMEOUT_S"
RAY_USAGE_STATS_KUBERAY_IN_USE = "RAY_USAGE_STATS_KUBERAY_IN_USE"
RAY_USAGE_STATS_EXTRA_TAGS = "RAY_USAGE_STATS_EXTRA_TAGS"
RAYCLUSTER_DEFAULT_REQUEUE_SECONDS_ENV = "RAYCLUSTER_DEFAULT_REQUEUE_SECONDS_ENV"
RAYCLUSTER_DEFAULT_REQUEUE_SECONDS = 300
KUBERAY_GEN_RAY_START_CMD = "KUBERAY_GEN_RAY_START_CMD"
// Environment variables for RayJob submitter Kubernetes Job.
// Example: ray job submit --address=http://$RAY_DASHBOARD_ADDRESS --submission-id=$RAY_JOB_SUBMISSION_ID ...
RAY_DASHBOARD_ADDRESS = "RAY_DASHBOARD_ADDRESS"
RAY_JOB_SUBMISSION_ID = "RAY_JOB_SUBMISSION_ID"
// Environment variables for Ray Autoscaler V2.
// The value of RAY_CLOUD_INSTANCE_ID is the Pod name for Autoscaler V2 alpha. This may change in the future.
RAY_CLOUD_INSTANCE_ID = "RAY_CLOUD_INSTANCE_ID"
// 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_NODE_TYPE_NAME = "RAY_NODE_TYPE_NAME"
// This KubeRay operator environment variable is used to determine if random Pod
// deletion should be enabled. Note that this only takes effect when autoscaling
// is enabled for the RayCluster. This is a feature flag for v0.6.0, and will be
// removed if the default behavior is stable enoguh.
ENABLE_RANDOM_POD_DELETE = "ENABLE_RANDOM_POD_DELETE"
// This KubeRay operator environment variable is used to determine if the Redis
// cleanup Job should be enabled. This is a feature flag for v1.0.0.
ENABLE_GCS_FT_REDIS_CLEANUP = "ENABLE_GCS_FT_REDIS_CLEANUP"
// 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.
// Enabling this feature contributes to the robustness of Ray clusters. It is currently a feature
// flag for v1.1.0 and will be removed if the behavior proves to be stable enough.
ENABLE_PROBES_INJECTION = "ENABLE_PROBES_INJECTION"
// Ray core default configurations
DefaultWorkerRayGcsReconnectTimeoutS = "600"
LOCAL_HOST = "127.0.0.1"
// Ray FT default readiness probe values
DefaultReadinessProbeInitialDelaySeconds = 10
DefaultReadinessProbeTimeoutSeconds = 1
DefaultReadinessProbePeriodSeconds = 5
DefaultReadinessProbeSuccessThreshold = 1
DefaultReadinessProbeFailureThreshold = 10
ServeReadinessProbeFailureThreshold = 1
// Ray FT default liveness probe values
DefaultLivenessProbeInitialDelaySeconds = 30
DefaultLivenessProbeTimeoutSeconds = 1
DefaultLivenessProbePeriodSeconds = 5
DefaultLivenessProbeSuccessThreshold = 1
DefaultLivenessProbeFailureThreshold = 120
// 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. So, RayAgentRayletHealthPath accesses the dashboard agent's API endpoint
// to check the health of the Raylet process.
// TODO (kevin85421): Should we take the dashboard process into account?
RayAgentRayletHealthPath = "api/local_raylet_healthz"
RayDashboardGCSHealthPath = "api/gcs_healthz"
RayServeProxyHealthPath = "-/healthz"
BaseWgetHealthCommand = "wget -T 2 -q -O- http://localhost:%d/%s | grep success"
// Finalizers for RayJob
RayJobStopJobFinalizer = "ray.io/rayjob-finalizer"
// RayNodeHeadGroupLabelValue is the value for the RayNodeGroupLabelKey label on a head node
RayNodeHeadGroupLabelValue = "headgroup"
// Telemetry
KUBERAY_VERSION = "v1.1.1"
)
type ServiceType string
const (
HeadService ServiceType = "headService"
ServingService ServiceType = "serveService"
)
// 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.
func RayOriginatedFromCRDLabelValue(crdType CRDType) string {
return string(crdType)
}