apache/beam · error · ValueError

A VPC network must be provided to use a private endpoint.

Error message

A VPC network must be provided to use a private endpoint.

What it means

VertexAIModelHandler raises ValueError in __init__ when private=True (a private Google Access endpoint) is requested but no VPC network is supplied. Private endpoints can only be reached through a specific VPC network, so the network parameter is mandatory in that mode.

Solutions

Pass the full VPC network resource name, e.g. network='projects/PROJECT/global/networks/NETWORK', when setting private=True
Set private=False if you actually intend to use the public endpoint
Verify the network exists in the same project and that the caller has network permissions

Example fix

// before
handler = VertexAIModelHandlerGPU(endpoint_id=ep, project=p, location=l, private=True)
// after
handler = VertexAIModelHandlerGPU(endpoint_id=ep, project=p, location=l, private=True, network='projects/p/global/networks/my-vpc')

Defensive patterns

Strategy: validation

Validate before calling

private = True
network = os.environ.get('VPC_NETWORK')
if private and not network:
    raise ValueError('private=True requires network="projects/P/global/networks/N"')
handler = VertexAIModelHandlerGPU(endpoint_id=ep, project=p, location=l, private=private, network=network)

Try / catch

try:
    handler = VertexAIModelHandlerGPU(..., private=True, network=network)
except ValueError as e:
    logger.error('Vertex handler config invalid: %s', e)
    raise

Prevention

Always set network together with private=True
Store the VPC network resource name in pipeline config/env, not hardcoded flags
Document that private endpoints require VPC access from the workers

When it happens

Trigger: Constructing VertexAIModelHandlerGPU with private=True while network=None (or network omitted).

Common situations: Deploying inference in a VPC-SC or private Google Access environment; copying public-endpoint config and only toggling private=True.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/85cdfd790348610e. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/apache_beam/ml/inference/vertex_ai_inference.py:144

    self._env_vars = kwargs.get('env_vars', {})
    self._invoke_route = invoke_route
    if min_batch_size is not None:
      self._batching_kwargs["min_batch_size"] = min_batch_size
    if max_batch_size is not None:
      self._batching_kwargs["max_batch_size"] = max_batch_size
    if max_batch_duration_secs is not None:
      self._batching_kwargs["max_batch_duration_secs"] = max_batch_duration_secs
    if max_batch_weight is not None:
      self._batching_kwargs["max_batch_weight"] = max_batch_weight
    if element_size_fn is not None:
      self._batching_kwargs['element_size_fn'] = element_size_fn
    if batch_length_fn is not None:
      self._batching_kwargs['length_fn'] = batch_length_fn
    if batch_bucket_boundaries is not None:
      self._batching_kwargs['bucket_boundaries'] = batch_bucket_boundaries

    if private and network is None:
      raise ValueError(
          "A VPC network must be provided to use a private endpoint.")

    # TODO: support the full list of options for aiplatform.init()
    # See https://cloud.google.com/python/docs/reference/aiplatform/latest/google.cloud.aiplatform#google_cloud_aiplatform_init
    aiplatform.init(
        project=project,
        location=location,
        experiment=experiment,
        network=network)

    # Check for liveness here but don't try to actually store the endpoint
    # in the class yet
    self.endpoint_name = endpoint_id
    self.location = location
    self.is_private = private

    _ = self._retrieve_endpoint(
        self.endpoint_name, self.location, self.is_private)

View on GitHub (pinned to 12126d8942)