GetLogs

Returns the logs for a Kubeflow job by initializing a log plugin and retrieving logs for master, worker, parameter server, chief, and evaluator replicas based on the task type and replica counts.

def GetLogs(
    pluginContext: k8s.PluginContext,
    taskType: string,
    objectMeta: meta_v1.ObjectMeta,
    taskTemplate: *core.TaskTemplate,
    hasMaster: bool,
    workersCount: int32,
    psReplicasCount: int32,
    chiefReplicasCount: int32,
    evaluatorReplicasCount: int32,
    primaryContainerName: string
) - > []*core.TaskLog, error

Retrieves the logs for a Kubeflow job, including logs for master, worker, parameter server, chief, and evaluator replicas based on the task type and replica counts. This function aggregates logs from various components of a distributed Kubeflow job.

Parameters

Name	Type	Description
pluginContext	k8s.PluginContext	The context for the Kubernetes plugin, providing access to task execution metadata.
taskType	string	The type of the task (e.g., PytorchTaskType, MPITaskType), which determines which specific logs to retrieve.
objectMeta	meta_v1.ObjectMeta	Kubernetes object metadata, containing the name and namespace of the Kubeflow job to identify the pods.
taskTemplate	*core.TaskTemplate	The task template associated with the job, used for initializing the log plugin.
hasMaster	bool	A boolean indicating whether the job has a master replica, relevant for Pytorch tasks.
workersCount	int32	The number of worker replicas in the job, used to iterate and retrieve logs for each worker.
psReplicasCount	int32	The number of parameter server replicas in the job, used to retrieve logs for each parameter server.
chiefReplicasCount	int32	The number of chief replicas in the job (typically 0 or 1), used to retrieve logs for the chief replica.
evaluatorReplicasCount	int32	The number of evaluator replicas in the job (typically 0 or 1), used to retrieve logs for the evaluator replica.
primaryContainerName	string	The name of the primary container within the pods from which to fetch logs.

Returns

Type	Description
[]*core.TaskLog, error	A slice of TaskLog objects containing the aggregated logs, or an error if log retrieval fails for any component.