Function TEST_F

src/tests/containerizer/nvidia_gpu_isolator_tests.cpp:93–191

This test verifies that we are able to enable the Nvidia GPU isolator and launch tasks with restricted access to GPUs. We first launch a task with access to 0 GPUs and verify that a call to `nvidia-smi` fails. We then launch a task with 1 GPU and verify that a call to `nvidia-smi` both succeeds and reports exactly 1 GPU available.

// and verify that a call to `nvidia-smi` both succeeds and
// reports exactly 1 GPU available.
TEST_F(NvidiaGpuTest, ROOT_CGROUPS_NVIDIA_GPU_VerifyDeviceAccess)
{
  Try<Owned<cluster::Master>> master = StartMaster();
  ASSERT_SOME(master);

  // Turn on Nvidia GPU isolation.
  // Assume at least one GPU is available for isolation.
  slave::Flags flags = CreateSlaveFlags();
  flags.isolation = "filesystem/linux,cgroups/devices,gpu/nvidia";
  flags.resources = "cpus:1"; // To override the default with gpus:0.

  Owned<MasterDetector> detector = master.get()->createDetector();

  Try<Owned<cluster::Slave>> slave = StartSlave(detector.get(), flags);
  ASSERT_SOME(slave);

  MockScheduler sched;

  FrameworkInfo frameworkInfo = DEFAULT_FRAMEWORK_INFO;
  frameworkInfo.add_capabilities()->set_type(
      FrameworkInfo::Capability::GPU_RESOURCES);

  MesosSchedulerDriver driver(
      &sched, frameworkInfo, master.get()->pid, DEFAULT_CREDENTIAL);

  Future<Nothing> schedRegistered;
  EXPECT_CALL(sched, registered(_, _, _))
    .WillOnce(FutureSatisfy(&schedRegistered));

  Future<vector<Offer>> offers1, offers2;
  EXPECT_CALL(sched, resourceOffers(_, _))
    .WillOnce(FutureArg<1>(&offers1))
    .WillOnce(FutureArg<1>(&offers2))
    .WillRepeatedly(Return());      // Ignore subsequent offers.

  driver.start();

  AWAIT_READY(schedRegistered);

  // Launch a task requesting no GPUs and
  // verify that running `nvidia-smi` fails.
  AWAIT_READY(offers1);
  EXPECT_EQ(1u, offers1->size());

  TaskInfo task1 = createTask(
      offers1.get()[0].slave_id(),
      Resources::parse("cpus:0.1;mem:128").get(),
      "nvidia-smi");

  Future<TaskStatus> statusStarting1, statusRunning1, statusFailed1;
  EXPECT_CALL(sched, statusUpdate(_, _))
    .WillOnce(FutureArg<1>(&statusStarting1))
    .WillOnce(FutureArg<1>(&statusRunning1))
    .WillOnce(FutureArg<1>(&statusFailed1));

  driver.launchTasks(offers1.get()[0].id(), {task1});

  AWAIT_READY(statusStarting1);