codeflare_sdk.vendored.python_client.utils package

Submodules

codeflare_sdk.vendored.python_client.utils.kuberay_cluster_builder module

Set of helper methods to manage rayclusters. Requires Python 3.9 and higher

class codeflare_sdk.vendored.python_client.utils.kuberay_cluster_builder.ClusterBuilder

Bases: IClusterBuilder

ClusterBuilder implements the abstract methods of IClusterBuilder to build a cluster.

build_head(ray_image: str = 'rayproject/ray:2.46.0', service_type: str = 'ClusterIP', cpu_requests: str = '2', memory_requests: str = '3G', cpu_limits: str = '2', memory_limits: str = '3G', ray_start_params: dict = {'dashboard-host': '0.0.0.0'})

Build head node of the ray cluster.

Parameters: - ray_image (str): Docker image for the head node. Default value is “rayproject/ray:2.46.0”. - service_type (str): Service type of the head node. Default value is “ClusterIP”, which creates a headless ClusterIP service. - cpu_requests (str): CPU requests for the head node. Default value is “2”. - memory_requests (str): Memory requests for the head node. Default value is “3G”. - cpu_limits (str): CPU limits for the head node. Default value is “2”. - memory_limits (str): Memory limits for the head node. Default value is “3G”. - ray_start_params (dict): Dictionary of start parameters for the head node. Default values is “dashboard-host”: “0.0.0.0”.

build_meta(name: str, k8s_namespace: str = 'default', labels: dict = None, ray_version: str = '2.46.0')

Builds the metadata and ray version of the cluster.

Parameters: - name (str): The name of the cluster. - k8s_namespace (str, optional): The namespace in which the Ray cluster exists. Defaults to “default”. - labels (dict, optional): A dictionary of key-value pairs to add as labels to the cluster. Defaults to None. - ray_version (str, optional): The version of Ray to use for the cluster. Defaults to “2.46.0”.

build_worker(group_name: str, ray_image: str = 'rayproject/ray:2.46.0', ray_command: Any = ['/bin/bash', '-lc'], init_image: str = 'busybox:1.28', cpu_requests: str = '1', memory_requests: str = '1G', cpu_limits: str = '2', memory_limits: str = '2G', replicas: int = 1, min_replicas: int = -1, max_replicas: int = -1, ray_start_params: dict = {})

Build worker specifications of the cluster.

This function sets the worker configuration of the cluster, including the Docker image, CPU and memory requirements, number of replicas, and other parameters.

Parameters: - group_name (str): name of the worker group. - ray_image (str, optional): Docker image for the Ray process. Default is “rayproject/ray:2.46.0”. - ray_command (Any, optional): Command to run in the Docker container. Default is [“/bin/bash”, “-lc”]. - init_image (str, optional): Docker image for the init container. Default is “busybox:1.28”. - cpu_requests (str, optional): CPU requests for the worker pods. Default is “1”. - memory_requests (str, optional): Memory requests for the worker pods. Default is “1G”. - cpu_limits (str, optional): CPU limits for the worker pods. Default is “2”. - memory_limits (str, optional): Memory limits for the worker pods. Default is “2G”. - replicas (int, optional): Number of worker pods to run. Default is 1. - min_replicas (int, optional): Minimum number of worker pods to run. Default is -1. - max_replicas (int, optional): Maximum number of worker pods to run. Default is -1. - ray_start_params (dict, optional): Additional parameters to pass to the ray start command. Default is {}.

get_cluster()

Returns the built cluster

class codeflare_sdk.vendored.python_client.utils.kuberay_cluster_builder.Director

Bases: object

build_basic_cluster(name: str, k8s_namespace: str = 'default', labels: dict = None) → dict

Builds a basic cluster with the given name and k8s_namespace parameters.

Parameters: - name (str): The name of the cluster. - k8s_namespace (str, optional): The kubernetes namespace for the cluster, with a default value of “default”.

Returns: dict: The basic cluster as a dictionary.

build_large_cluster(name: str, k8s_namespace: str = 'default', labels: dict = None) → dict

Builds a medium cluster with the given name and k8s_namespace parameters. with 1 workergroup, the workgroup has 6 replicas with 6 cpu and 6G memory limits

Parameters: - name (str): The name of the cluster. - k8s_namespace (str, optional): The kubernetes namespace for the cluster, with a default value of “default”.

Returns: dict: The small cluster as a dictionary.

build_medium_cluster(name: str, k8s_namespace: str = 'default', labels: str = None) → dict

Builds a medium cluster with the given name and k8s_namespace parameters with 1 workergroup, the workgroup has 3 replicas with 4 cpu and 4G memory limits

Parameters: - name (str): The name of the cluster. - k8s_namespace (str, optional): The kubernetes namespace for the cluster, with a default value of “default”.

Returns: dict: The small cluster as a dictionary.

build_small_cluster(name: str, k8s_namespace: str = 'default', labels: dict = None) → dict

Builds a small cluster with the given name and k8s_namespace parameters with 1 workergroup, the workgroup has 1 replica with 2 cpu and 2G memory limits

Parameters: - name (str): The name of the cluster. - k8s_namespace (str, optional): The kubernetes namespace for the cluster, with a default value of “default”.

Returns: dict: The small cluster as a dictionary.

class codeflare_sdk.vendored.python_client.utils.kuberay_cluster_builder.IClusterBuilder

Bases: object

IClusterBuilder is an interface for building a cluster.

The class defines abstract methods for building the metadata, head pod, worker groups, and retrieving the built cluster.

abstract static build_head()

builds the head pod

abstract static build_meta()

builds the cluster metadata

abstract static build_worker()

builds a worker group

abstract static get_cluster()

Returns the built cluster

codeflare_sdk.vendored.python_client.utils.kuberay_cluster_utils module

Set of helper methods to manage rayclusters. Requires Python 3.6 and higher

class codeflare_sdk.vendored.python_client.utils.kuberay_cluster_utils.ClusterUtils

Bases: object

ClusterUtils - Utility class for populating cluster information

Methods: - populate_meta(cluster: dict, name: str, k8s_namespace: str, labels: dict, ray_version: str) -> dict: - populate_ray_head(cluster: dict, ray_image: str,service_type: str, cpu_requests: str, memory_requests: str, cpu_limits: str, memory_limits: str, ray_start_params: dict) -> Tuple[dict, bool]: - populate_worker_group(cluster: dict, group_name: str, ray_image: str, ray_command: Any, init_image: str, cpu_requests: str, memory_requests: str, cpu_limits: str, memory_limits: str, replicas: int, min_replicas: int, max_replicas: int, ray_start_params: dict) -> Tuple[dict, bool]: - update_worker_group_replicas(cluster: dict, group_name: str, max_replicas: int, min_replicas: int, replicas: int) -> Tuple[dict, bool]:

delete_worker_group(cluster: dict, group_name: str) → Tuple[dict, bool]

Deletes a worker group in the cluster.

Parameters: - cluster (dict): The cluster definition. - group_name (str): The name of the worker group to be duplicated.

Returns: Tuple[dict, bool]: A tuple containing the updated cluster definition and a boolean indicating the success of the operation.

duplicate_worker_group(cluster: dict, group_name: str, new_group_name: str) → Tuple[dict, bool]

Duplicate a worker group in the cluster.

Parameters: - cluster (dict): The cluster definition. - group_name (str): The name of the worker group to be duplicated. - new_group_name (str): The name for the duplicated worker group.

Returns: Tuple[dict, bool]: A tuple containing the updated cluster definition and a boolean indicating the success of the operation.

is_valid_label(name: str) → bool

is_valid_name(name: str) → bool

populate_meta(cluster: dict, name: str, k8s_namespace: str, labels: dict, ray_version: str) → dict

Populate the metadata and ray version of the cluster.

Parameters: - cluster (dict): A dictionary representing a cluster. - name (str): The name of the cluster. - k8s_namespace (str): The namespace of the cluster. - labels (dict): A dictionary of labels to be applied to the cluster. - ray_version (str): The version of Ray to use in the cluster.

Returns:

dict: The updated cluster dictionary with metadata and ray version populated.

populate_ray_head(cluster: dict, ray_image: str, service_type: str, cpu_requests: str, memory_requests: str, cpu_limits: str, memory_limits: str, ray_start_params: dict) → Tuple[dict, bool]

Populate the ray head specs of the cluster Parameters: - cluster (dict): The dictionary representation of the cluster. - ray_image (str): The name of the ray image to use for the head node. - service_type (str): The type of service to run for the head node. - cpu_requests (str): The CPU resource requests for the head node. - memory_requests (str): The memory resource requests for the head node. - cpu_limits (str): The CPU resource limits for the head node. - memory_limits (str): The memory resource limits for the head node. - ray_start_params (dict): The parameters for starting the Ray cluster.

Returns: - Tuple (dict, bool): The updated cluster, and a boolean indicating whether the update was successful.

populate_worker_group(group_name: str, ray_image: str, ray_command: Any, init_image: str, cpu_requests: str, memory_requests: str, cpu_limits: str, memory_limits: str, replicas: int, min_replicas: int, max_replicas: int, ray_start_params: dict) → Tuple[dict, bool]

Populate the worker group specification in the cluster dictionary.

Parameters: - cluster (dict): Dictionary representing the cluster spec. - group_name (str): The name of the worker group. - ray_image (str): The image to use for the Ray worker containers. - ray_command (Any): The command to run in the Ray worker containers. - init_image (str): The init container image to use. - cpu_requests (str): The requested CPU resources for the worker containers. - memory_requests (str): The requested memory resources for the worker containers. - cpu_limits (str): The limit on CPU resources for the worker containers. - memory_limits (str): The limit on memory resources for the worker containers. - replicas (int): The desired number of replicas for the worker group. - min_replicas (int): The minimum number of replicas for the worker group. - max_replicas (int): The maximum number of replicas for the worker group. - ray_start_params (dict): The parameters to pass to the Ray worker start command.

Returns: - Tuple[dict, bool]: A tuple of the cluster specification and a boolean indicating

whether the worker group was successfully populated.

update_worker_group_replicas(cluster: dict, group_name: str, max_replicas: int, min_replicas: int, replicas: int) → Tuple[dict, bool]

Update the number of replicas for a worker group in the cluster.

Parameters: - cluster (dict): The cluster to update. - group_name (str): The name of the worker group to update. - max_replicas (int): The maximum number of replicas for the worker group. - min_replicas (int): The minimum number of replicas for the worker group. - replicas (int): The desired number of replicas for the worker group.

Returns: Tuple[dict, bool]: A tuple containing the updated cluster and a flag indicating whether the update was successful.

update_worker_group_resources(cluster: dict, group_name: str, cpu_requests: str, memory_requests: str, cpu_limits: str, memory_limits: str, container_name='unspecified') → Tuple[dict, bool]

Update the resources for a worker group pods in the cluster.

Parameters: - cluster (dict): The cluster to update. - group_name (str): The name of the worker group to update. - cpu_requests (str): CPU requests for the worker pods. - memory_requests (str): Memory requests for the worker pods. - cpu_limits (str): CPU limits for the worker pods. - memory_limits (str): Memory limits for the worker pods.

Returns: Tuple[dict, bool]: A tuple containing the updated cluster and a flag indicating whether the update was successful.

Module contents