TFE_ExecutorWaitForAllPendingNodes(executor, status);
  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);

  // TODO(b/136478427): Figure out how to correctly shut the server down.
  worker_server.release();

  // Update the server def with a new set of names (worker instead of
  // localhost).
  tensorflow::ServerDef updated_server_def = GetServerDef("worker", 2);

    ],
)

tf_cuda_cc_test(
    name = "c_api_remote_test",
    size = "medium",
    srcs = [
        "c_api_remote_test.cc",
    ],
    # TODO(b/136478427): Figure out how to correctly shut the server down
    args = ["--heap_check="],
    tags = [
        "no_windows",
    ],
    deps = [
        ":c_api",
        ":c_api_experimental",
        ":c_api_internal",
        ":c_api_remote_test_util",

  TFE_ContextRemoveFunction(ctx, "AddVariablesFunction", status);
  TFE_DeleteContext(ctx);

  TF_DeleteStatus(status);

  // TODO(b/136478427): Figure out how to correctly shut the server down.
  worker_server1.release();
  worker_server2.release();
}

TEST(CAPI, TestLocalFunctionWithPackedInput) {
  TestFunctionWithPackedInput(/*remote=*/false);
}

    ```

    **Note:** You can add `--test_sharding_strategy=disabled` to the `flags` to
    disable the sharding so that all the test outputs are in one file. However,
    it may slow down the tests for not running in parallel and may cause the
    test to timeout but it could be useful when you need to execute a single
    test or more in general your filtered/selected tests have a very low

      // One of the inputs we're trying to pack is on a custom device. We'll let
      // the first custom device we see handle all of the packing.
      auto* custom_device_handle =
          tensorflow::down_cast<tensorflow::CustomDeviceTensorHandle*>(
              unwrapped_handle);
      tensorflow::ImmediateExecutionTensorHandle* result;
      status->status = custom_device_handle->device()->Pack(

    if (!t) {
      return tensorflow::errors::InvalidArgument(
          "Unable to cast input to GraphTensor");
    }
    inputs_.push_back(t->output_);
    *output = tensorflow::down_cast<TracingTensorHandle*>(outputs[0]);
    return absl::OkStatus();
  }

  Status Finalize(OutputList* outputs, AbstractFunction** f) override {
    std::vector<TF_Output> graph_outputs;

// Checks the health of collective ops peers. Explicit health check is needed in
// multi worker collective ops to detect failures in the cluster.  If a peer is
// down, collective ops may hang.
TF_CAPI_EXPORT extern void TFE_CollectiveOpsCheckPeerHealth(
    TFE_Context* ctx, const char* task, int64_t timeout_in_ms,
    TF_Status* status);

  TF_AbstractFunction* func =
      TF_FinalizeFunction(graph_ctx, add_outputs, status.get());
  ASSERT_EQ(TF_OK, TF_GetCode(status.get())) << TF_Message(status.get());
  // Note: TF_OutputList does not own the underlying AbstractTensors, those
  // need to be deleted explicitly.
  TF_DeleteAbstractTensor(TF_OutputListGet(add_outputs, 0));

  // Build eager context.
  TFE_ContextOptions* opts = TFE_NewContextOptions();

// The config key must be set before getting its value. Getting value of
// non-existing config keys will result in errors.
// If `timeout_in_ms=0`, this call will block until the key-value is set or the
// worker shuts down.
TF_CAPI_EXPORT extern void TFE_GetConfigKeyValue(TFE_Context* ctx,
                                                 const char* key,
                                                 int64_t timeout_in_ms,

      tensorflow::unwrap(h);
  // TODO(b/175427838): It would be nice to be able to use tensorflow::isa here.
  if (tensorflow::CustomDeviceTensorHandle::classof(unwrapped_handle)) {
    return tensorflow::down_cast<tensorflow::CustomDeviceTensorHandle*>(
               unwrapped_handle)
        ->DevicePointer();
  }
  // TODO(b/175427838): It would be nice to be able to use tensorflow::isa here.