executed = false;
  TFE_Execute(op.get(), &var_handle, &num_retvals, status.get());
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
  ASSERT_TRUE(executed);
  auto handle_cleaner = tensorflow::gtl::MakeCleanup(
      [var_handle]() { TFE_DeleteTensorHandle(var_handle); });

  // Assign to the variable, copying to the custom device.

  // caller of event_mgr->ThenExecute() blocks on the completion of the callback
  // (as in the case of ConstOp kernel creation on GPU, which involves copying a
  // CPU tensor to GPU).
  // Setting a larger thread pool does not help with the Swift caller, as we use
  // a different TFE context for each thread of execution (for running graph
  // functions, and their send/recvs corountines).

    EXPECT_EQ(got_n, want_n);
    calls++;
    memset(buffer, 'x', got_n);
    TF_SetStatus(status, TF_OK, "");
    return got_n;
  };
  // If block_size, max_bytes, or both are zero, or want_n is larger than
  // max_bytes the cache is a pass-through.
  tf_gcs_filesystem::RamFileBlockCache cache1(1, 0, 0, fetcher);
  tf_gcs_filesystem::RamFileBlockCache cache2(0, 1, 0, fetcher);

    job_def->mutable_tasks()->insert(
        {i, tensorflow::strings::StrCat("localhost:", port)});
  }
  auto* config = server_def.mutable_default_session_config();
  config->mutable_experimental()->set_collective_group_leader(
      tensorflow::strings::StrCat("/job:", job_name, "/replica:0/task:", 0));
  auto* rewrite_options =
      config->mutable_graph_options()->mutable_rewrite_options();
  rewrite_options->set_scoped_allocator_optimization(