// Create two contexts.
  int32_t init_timeout_in_ms = 300000;
  TFE_Context* ctx_0 =
      CreateContext(serialized_server_def_0,
                    /*isolate_session_state=*/false, init_timeout_in_ms);
  TFE_Context* ctx_1 =
      CreateContext(serialized_server_def_1,
                    /*isolate_session_state=*/false, init_timeout_in_ms);

  // Remote device on `worker2`.

        "Coordination service is not enabled.");
    return;
  }
  absl::Duration timeout;
  if (timeout_in_ms > 0) {
    timeout = absl::Milliseconds(timeout_in_ms);
  } else {
    // Block until the key-value is set or the worker shuts down.
    timeout = absl::InfiniteDuration();
  }
  auto status_or_value = coord_agent->GetKeyValue(key, timeout);
  status->status = status_or_value.status();
  if (!status_or_value.ok()) return;

    TFE_Context* ctx, int keep_alive_secs, const void* proto, size_t proto_len,
    int64_t init_timeout_in_ms, TF_Status* status);

// This API is for experimental usage and may be subject to change.
TF_CAPI_EXPORT extern void TFE_ContextSetServerDefWithTimeout(
    TFE_Context* ctx, int keep_alive_secs, const void* proto, size_t proto_len,
    int64_t init_timeout_in_ms, TF_Status* status,
    bool clear_existing_contexts);

      ctx, keep_alive_secs, proto, proto_len, /*init_timeout_in_ms=*/0,
      /*retries=*/0, status, /*clear_existing_contexts=*/false);
}

// Set server def with timeout.
TF_CAPI_EXPORT extern void TFE_ContextSetServerDefWithTimeout(
    TFE_Context* ctx, int keep_alive_secs, const void* proto, size_t proto_len,
    int64_t init_timeout_in_ms, TF_Status* status,
    bool clear_existing_contexts) {

// Caller must call TF_DeleteTensor() over the returned tensor. If the queue is
// empty, this call is blocked.
//
// Tensors are enqueued via the corresponding TF enqueue op.
// TODO(hongm): Add support for `timeout_ms`.
TF_CAPI_EXPORT extern TF_Tensor* TF_DequeueNamedTensor(TF_Session* session,
                                                       int tensor_id,
                                                       TF_Status* status);

TF_Buffer* TF_CoordinationServiceGetKeyValueWithTimeout(
    const char* key, int64_t key_size, int64_t timeout_seconds,
    TF_CoordinationServiceAgent* agent, TF_Status* status) {
  if (timeout_seconds <= 0) {
    status->status = absl::InvalidArgumentError(
        "TF_CoordinationServiceGetKeyValueWithTimeout invoked with invalid "
        "timeout_seconds <= 0.");
    return nullptr;
  }

  TestRemoteExecuteUpdateServerDefResourceAccess(true);
}

void TestRemoteExecuteUpdateServerDefWithFailures(bool async) {
  // Fail fast on GetStatus requests so we can get errors instead of timeout
  // when updating cluster with non-exsitent worker
  tensorflow::setenv("GRPC_FAIL_FAST", "TRUE", /*overwrite=*/1);

  tensorflow::ServerDef server_def = GetServerDef(2);

  TF_DeleteStatus(status);
  return var_handle;
}

TFE_Context* CreateContext(const std::string& serialized_server_def,
                           bool isolate_session_state,
                           int64_t init_timeout_in_ms) {
  TF_Status* status = TF_NewStatus();
  TFE_ContextOptions* opts = TFE_NewContextOptions();
  opts->session_options.options.config.set_isolate_session_state(
      isolate_session_state);

    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
    execution time and the sharding

    TFE_Context* ctx, const char* task, int64_t timeout_in_ms,
    TF_Status* status) {
  tensorflow::EagerContext* context =
      tensorflow::ContextFromInterface(tensorflow::unwrap(ctx));
  auto collective_executor_handle = context->GetCollectiveExecutorHandle();
  tensorflow::Notification done;
  collective_executor_handle->get()->remote_access()->CheckPeerHealth(
      task, timeout_in_ms, [&done, status](const Status& s) {