Skip to content

create_dataflow_endpoint_group_v2

create_dataflow_endpoint_group_v2(
    endpoints: CreateEndpointDetailsList,
    *,
    config_overrides: Optional[GroundStationClientConfig] = None,
    contact_pre_pass_duration_seconds: Optional[
        DataflowEndpointGroupDurationInSeconds
    ] = None,
    contact_post_pass_duration_seconds: Optional[
        DataflowEndpointGroupDurationInSeconds
    ] = None,
    tags: Optional[TagsMap] = None,
) -> CreateDataflowEndpointGroupV2Response

Creates a DataflowEndpoint group containing the specified list of Ground Station Agent based endpoints.

The name field in each endpoint is used in your mission profile DataflowEndpointConfig to specify which endpoints to use during a contact.

When a contact uses multiple DataflowEndpointConfig objects, each Config must match a DataflowEndpoint in the same group.

Parameters:

Raises:

Source code in docs_src/capo_groundstation/_services/ground_station.py
def create_dataflow_endpoint_group_v2(
    self,
    endpoints: "capo_groundstation.types.create_endpoint_details_list.CreateEndpointDetailsList",
    *,
    config_overrides: Optional[GroundStationClientConfig] = None,
    contact_pre_pass_duration_seconds: Optional[
        "capo_groundstation.types.dataflow_endpoint_group_duration_in_seconds.DataflowEndpointGroupDurationInSeconds"
    ] = None,
    contact_post_pass_duration_seconds: Optional[
        "capo_groundstation.types.dataflow_endpoint_group_duration_in_seconds.DataflowEndpointGroupDurationInSeconds"
    ] = None,
    tags: Optional["capo_groundstation.types.tags_map.TagsMap"] = None,
) -> "capo_groundstation.types.create_dataflow_endpoint_group_v2_response.CreateDataflowEndpointGroupV2Response":
    """<p>Creates a <code>DataflowEndpoint</code> group containing the specified list of Ground Station Agent based endpoints.</p> <p>The <code>name</code> field in each endpoint is used in your mission profile <code> DataflowEndpointConfig</code> to specify which endpoints to use during a contact.</p> <p>When a contact uses multiple <code>DataflowEndpointConfig</code> objects, each <code> Config</code> must match a <code>DataflowEndpoint</code> in the same group.</p>

    Args:
        endpoints: <p>Dataflow endpoint group's endpoint definitions</p>
        contact_pre_pass_duration_seconds: <p> Amount of time, in seconds, before a contact starts that the Ground Station Dataflow Endpoint Group will be in a <code>PREPASS</code> state. A <a href="https://docs.aws.amazon.com/ground-station/latest/ug/monitoring.automating-events.html">Ground Station Dataflow Endpoint Group State Change event</a> will be emitted when the Dataflow Endpoint Group enters and exits the <code>PREPASS</code> state. </p>
        contact_post_pass_duration_seconds: <p> Amount of time, in seconds, after a contact ends that the Ground Station Dataflow Endpoint Group will be in a <code>POSTPASS</code> state. A <a href="https://docs.aws.amazon.com/ground-station/latest/ug/monitoring.automating-events.html">Ground Station Dataflow Endpoint Group State Change event</a> will be emitted when the Dataflow Endpoint Group enters and exits the <code>POSTPASS</code> state. </p>
        tags: <p>Tags of a V2 dataflow endpoint group.</p>

    Raises:
        capo_groundstation.errors.dependency_exception.DependencyException: <p>Dependency encountered an error.</p>
        capo_groundstation.errors.invalid_parameter_exception.InvalidParameterException: <p>One or more parameters are not valid.</p>
        capo_groundstation.errors.resource_not_found_exception.ResourceNotFoundException: <p>Resource was not found.</p>
        capo_groundstation.errors.service_quota_exceeded_exception.ServiceQuotaExceededException: <p>Request would cause a service quota to be exceeded.</p>
        capo_groundstation.errors.UnknownServiceError: The service returned an error code this client does not model.
    """

    def _handler(
        req: "OperationRequest[capo_groundstation.types.create_dataflow_endpoint_group_v2_request.CreateDataflowEndpointGroupV2Request]",
    ) -> OperationResponse[
        "capo_groundstation.types.create_dataflow_endpoint_group_v2_response.CreateDataflowEndpointGroupV2Response"
    ]:
        import capo_groundstation._operations.ground_station.create_dataflow_endpoint_group_v2

        output, http_response = (
            capo_groundstation._operations.ground_station.create_dataflow_endpoint_group_v2.create_dataflow_endpoint_group_v2(
                req.options, req.input
            )
        )
        return OperationResponse(output=output, response=http_response)

    interceptors_, options_ = self.operation_options(
        config_overrides, OperationLogging("CreateDataflowEndpointGroupV2")
    )
    input_: capo_groundstation.types.create_dataflow_endpoint_group_v2_request.CreateDataflowEndpointGroupV2Request = {
        "endpoints": endpoints
    }
    if contact_pre_pass_duration_seconds is not None:
        input_["contact_pre_pass_duration_seconds"] = (
            contact_pre_pass_duration_seconds
        )
    if contact_post_pass_duration_seconds is not None:
        input_["contact_post_pass_duration_seconds"] = (
            contact_post_pass_duration_seconds
        )
    if tags is not None:
        input_["tags"] = tags

    response = execute_pipeline(
        OperationRequest(input=input_, options=options_),
        handler=_handler,
        interceptors=list(interceptors_),
    )
    response.response.close()
    return response.output