def test_process_cgroup_metric_with_no_memory_cgroup_mounted(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
with patch("azurelinuxagent.common.cgroup.CpuCgroup._get_cpu_percent") as patch_get_cpu_percent:
with patch("azurelinuxagent.common.cgroup.MemoryCgroup.get_memory_usage") as patch_get_memory_usage:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active") as patch_is_active:
patch_is_active.return_value = True
patch_get_memory_usage.side_effect = Exception("File not found")
current_cpu = 30
patch_get_cpu_percent.return_value = current_cpu
poll_count = 1
for data_count in range(poll_count, 10):
CGroupsTelemetry.poll_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
self._assert_cgroup_metrics_equal(cpu_usage=[current_cpu] * data_count, memory_usage=[], max_memory_usage=[])
CGroupsTelemetry.report_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
self._assert_cgroup_metrics_equal([], [], [])
def test_telemetry_polling_to_generate_transient_logs_ioerror_permission_denied(self, mock_read_file,
patch_periodic_warn, *args):
num_extensions = 1
num_controllers = 2
is_active_check_per_controller = 2
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
self.assertEqual(0, patch_periodic_warn.call_count)
# Expecting logs to be present for different kind of errors
io_error_3 = IOError()
io_error_3.errno = errno.EPERM
mock_read_file.side_effect = io_error_3
poll_count = 1
expected_count_per_call = num_controllers + is_active_check_per_controller
# each collect per controller would generate a log statement, and each cgroup would invoke a
# is active check raising an exception
for data_count in range(poll_count, 10):
CGroupsTelemetry.poll_all_tracked()
self.assertEqual(poll_count * expected_count_per_call, patch_periodic_warn.call_count)
def test_cleanup_legacy_cgroups_should_disable_cgroups_when_the_daemon_was_added_to_the_legacy_cgroup_on_systemd(self, _):
# Set up a mock /var/run/waagent.pid file
daemon_pid = "42"
daemon_pid_file = os.path.join(self.tmp_dir, "waagent.pid")
fileutil.write_file(daemon_pid_file, daemon_pid + "\n")
# Set up old controller cgroups and add the daemon PID to them
CGroupsTools.create_legacy_agent_cgroup(self.cgroups_file_system_root, "cpu", daemon_pid)
CGroupsTools.create_legacy_agent_cgroup(self.cgroups_file_system_root, "memory", daemon_pid)
# Start tracking a couple of dummy cgroups
CGroupsTelemetry.track_cgroup(CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service", "cpu"))
CGroupsTelemetry.track_cgroup(CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service", "memory"))
cgroup_configurator = CGroupConfigurator.get_instance()
with patch("azurelinuxagent.common.cgroupconfigurator.add_event") as mock_add_event:
with patch("azurelinuxagent.common.cgroupapi.get_agent_pid_file_path", return_value=daemon_pid_file):
cgroup_configurator.cleanup_legacy_cgroups()
self.assertEquals(len(mock_add_event.call_args_list), 1)
_, kwargs = mock_add_event.call_args_list[0]
self.assertEquals(kwargs['op'], 'CGroupsCleanUp')
self.assertFalse(kwargs['is_success'])
self.assertEquals(
kwargs['message'],
"Failed to process legacy cgroups. Collection of resource usage data will be disabled. [CGroupsException] The daemon's PID ({0}) was already added to the legacy cgroup; this invalidates resource usage data.".format(daemon_pid))
self.assertFalse(cgroup_configurator.enabled())
self.assertEquals(len(CGroupsTelemetry._tracked), 0)
def _track_new_extension_cgroups(num_extensions):
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
def test_is_active_file_not_present(self, patch_periodic_warn):
test_cgroup = CGroup.create(os.path.join(data_dir, "cgroups", "not_cpu_mount"), "cpu", "test_extension")
self.assertEqual(False, test_cgroup.is_active())
test_cgroup = CGroup.create(os.path.join(data_dir, "cgroups", "not_memory_mount"), "memory", "test_extension")
self.assertEqual(False, test_cgroup.is_active())
self.assertEqual(0, patch_periodic_warn.call_count)
def test_is_active_incorrect_file(self, patch_periodic_warn):
test_cgroup = CGroup.create(os.path.join(data_dir, "cgroups", "cpu_mount", "tasks"), "cpu", "test_extension")
self.assertEqual(False, test_cgroup.is_active())
self.assertEqual(1, patch_periodic_warn.call_count)
test_cgroup = CGroup.create(os.path.join(data_dir, "cgroups", "memory_mount", "tasks"), "memory", "test_extension")
self.assertEqual(False, test_cgroup.is_active())
self.assertEqual(2, patch_periodic_warn.call_count)
def test_telemetry_polling_with_changing_cgroups_state(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
with patch("azurelinuxagent.common.cgroup.MemoryCgroup.get_max_memory_usage") as patch_get_memory_max_usage:
with patch("azurelinuxagent.common.cgroup.MemoryCgroup.get_memory_usage") as patch_get_memory_usage:
with patch("azurelinuxagent.common.cgroup.CpuCgroup._get_cpu_percent") as patch_get_cpu_percent:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active") as patch_is_active:
patch_is_active.return_value = True
no_extensions_expected = 0
expected_data_count = 2
current_cpu = 30
current_memory = 209715200
current_max_memory = 471859200
patch_get_cpu_percent.return_value = current_cpu
patch_get_memory_usage.return_value = current_memory # example 200 MB
patch_get_memory_max_usage.return_value = current_max_memory # example 450 MB
for i in range(num_extensions):
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
CGroupsTelemetry.poll_all_tracked()
for i in range(num_extensions):
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
patch_is_active.return_value = False
CGroupsTelemetry.poll_all_tracked()
for i in range(num_extensions):
self.assertFalse(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertFalse(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
self._assert_cgroup_metrics_equal(
cpu_usage=[current_cpu] * expected_data_count,
memory_usage=[current_memory] * expected_data_count,
max_memory_usage=[current_max_memory] * expected_data_count)
CGroupsTelemetry.report_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), no_extensions_expected)
self._assert_cgroup_metrics_equal([], [], [])
def test_telemetry_polling_with_inactive_cgroups(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
with patch("azurelinuxagent.common.cgroup.MemoryCgroup._get_memory_max_usage") as patch_get_memory_max_usage:
with patch("azurelinuxagent.common.cgroup.MemoryCgroup._get_memory_usage") as patch_get_memory_usage:
with patch("azurelinuxagent.common.cgroup.CpuCgroup._get_cpu_percent") as patch_get_cpu_percent:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active") as patch_is_active:
patch_is_active.return_value = False
no_extensions_expected = 0
data_count = 1
current_cpu = 30
current_memory = 209715200
current_max_memory = 471859200
patch_get_cpu_percent.return_value = current_cpu
patch_get_memory_usage.return_value = current_memory # example 200 MB
patch_get_memory_max_usage.return_value = current_max_memory # example 450 MB
for i in range(num_extensions):
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
CGroupsTelemetry.poll_all_tracked()
for i in range(num_extensions):
self.assertFalse(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertFalse(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
for cgroup_name, cgroup_metric in CGroupsTelemetry._cgroup_metrics.items():
current_memory_usage, max_memory_levels, current_cpu_usage = cgroup_metric.get_metrics()
self.assertEqual(len(current_memory_usage._data), data_count)
self.assertListEqual(current_memory_usage._data, [current_memory] * data_count)
self.assertEqual(len(max_memory_levels._data), data_count)
self.assertListEqual(max_memory_levels._data, [current_max_memory] * data_count)
self.assertEqual(len(current_cpu_usage._data), data_count)
self.assertListEqual(current_cpu_usage._data, [current_cpu] * data_count)
CGroupsTelemetry.report_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), no_extensions_expected)
for cgroup_name, cgroup_metric in CGroupsTelemetry._cgroup_metrics.items():
current_memory_usage, max_memory_levels, current_cpu_usage = cgroup_metric.get_metrics()
self.assertEqual(len(current_memory_usage._data), 0)
self.assertEqual(len(max_memory_levels._data), 0)
self.assertEqual(len(current_cpu_usage._data), 0)
def test_cleanup_legacy_cgroups_should_disable_cgroups_when_it_fails_to_process_legacy_cgroups(
self):
# Set up a mock /var/run/waagent.pid file
daemon_pid = "42"
daemon_pid_file = os.path.join(self.tmp_dir, "waagent.pid")
fileutil.write_file(daemon_pid_file, daemon_pid + "\n")
# Set up old controller cgroups and add the daemon PID to them
CGroupsTools.create_legacy_agent_cgroup(self.cgroups_file_system_root,
"cpu", daemon_pid)
CGroupsTools.create_legacy_agent_cgroup(self.cgroups_file_system_root,
"memory", daemon_pid)
# Set up new controller cgroups and add extension handler's PID to them
CGroupsTools.create_agent_cgroup(self.cgroups_file_system_root, "cpu",
"999")
CGroupsTools.create_agent_cgroup(self.cgroups_file_system_root,
"memory", "999")
def mock_append_file(filepath, contents, **kwargs):
if re.match(r'/.*/cpu/.*/cgroup.procs', filepath):
raise OSError(errno.ENOSPC, os.strerror(errno.ENOSPC))
fileutil.append_file(filepath, contents, **kwargs)
# Start tracking a couple of dummy cgroups
CGroupsTelemetry.track_cgroup(
CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service",
"cpu"))
CGroupsTelemetry.track_cgroup(
CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service",
"memory"))
cgroup_configurator = CGroupConfigurator.get_instance()
with patch("azurelinuxagent.common.cgroupconfigurator.add_event"
) as mock_add_event:
with patch(
"azurelinuxagent.common.cgroupapi.get_agent_pid_file_path",
return_value=daemon_pid_file):
with patch(
"azurelinuxagent.common.cgroupapi.fileutil.append_file",
side_effect=mock_append_file):
cgroup_configurator.cleanup_legacy_cgroups()
self.assertEquals(len(mock_add_event.call_args_list), 1)
_, kwargs = mock_add_event.call_args_list[0]
self.assertEquals(kwargs['op'], 'CGroupsCleanUp')
self.assertFalse(kwargs['is_success'])
self.assertEquals(
kwargs['message'],
'Failed to process legacy cgroups. Collection of resource usage data will be disabled. [Errno 28] No space left on device'
)
self.assertFalse(cgroup_configurator.enabled())
self.assertEquals(len(CGroupsTelemetry._tracked), 0)
def test_disable_should_reset_tracked_cgroups(self):
configurator = CGroupConfigurator.get_instance()
# Start tracking a couple of dummy cgroups
CGroupsTelemetry.track_cgroup(CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service", "cpu"))
CGroupsTelemetry.track_cgroup(CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service", "memory"))
configurator.disable()
self.assertFalse(configurator.enabled())
self.assertEquals(len(CGroupsTelemetry._tracked), 0)
def test_telemetry_calculations(self, *args):
num_polls = 10
num_extensions = 1
num_summarization_values = 7
cpu_percent_values = [random.randint(0, 100) for _ in range(num_polls)]
# only verifying calculations and not validity of the values.
memory_usage_values = [random.randint(0, 8 * 1024 ** 3) for _ in range(num_polls)]
max_memory_usage_values = [random.randint(0, 8 * 1024 ** 3) for _ in range(num_polls)]
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
self.assertEqual(2 * num_extensions, len(CGroupsTelemetry._tracked))
with patch("azurelinuxagent.common.cgroup.MemoryCgroup.get_max_memory_usage") as patch_get_memory_max_usage:
with patch("azurelinuxagent.common.cgroup.MemoryCgroup.get_memory_usage") as patch_get_memory_usage:
with patch("azurelinuxagent.common.cgroup.CpuCgroup._get_cpu_percent") as patch_get_cpu_percent:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active") as patch_is_active:
for i in range(num_polls):
patch_is_active.return_value = True
patch_get_cpu_percent.return_value = cpu_percent_values[i]
patch_get_memory_usage.return_value = memory_usage_values[i] # example 200 MB
patch_get_memory_max_usage.return_value = max_memory_usage_values[i] # example 450 MB
CGroupsTelemetry.poll_all_tracked()
collected_metrics = CGroupsTelemetry.report_all_tracked()
for i in range(num_extensions):
name = "dummy_extension_{0}".format(i)
self.assertIn(name, collected_metrics)
self.assertIn("memory", collected_metrics[name])
self.assertIn("cur_mem", collected_metrics[name]["memory"])
self.assertIn("max_mem", collected_metrics[name]["memory"])
self.assertEqual(num_summarization_values, len(collected_metrics[name]["memory"]["cur_mem"]))
self.assertEqual(num_summarization_values, len(collected_metrics[name]["memory"]["max_mem"]))
self.assertListEqual(generate_metric_list(memory_usage_values),
collected_metrics[name]["memory"]["cur_mem"][0:5])
self.assertListEqual(generate_metric_list(max_memory_usage_values),
collected_metrics[name]["memory"]["max_mem"][0:5])
self.assertIn("cpu", collected_metrics[name])
self.assertIn("cur_cpu", collected_metrics[name]["cpu"])
self.assertEqual(num_summarization_values, len(collected_metrics[name]["cpu"]["cur_cpu"]))
self.assertListEqual(generate_metric_list(cpu_percent_values),
collected_metrics[name]["cpu"]["cur_cpu"][0:5])
def test_correct_creation(self):
test_cgroup = CGroup.create("dummy_path", "cpu", "test_extension")
self.assertIsInstance(test_cgroup, CpuCgroup)
self.assertEqual(test_cgroup.controller, "cpu")
self.assertEqual(test_cgroup.path, "dummy_path")
self.assertEqual(test_cgroup.name, "test_extension")
test_cgroup = CGroup.create("dummy_path", "memory", "test_extension")
self.assertIsInstance(test_cgroup, MemoryCgroup)
self.assertEqual(test_cgroup.controller, "memory")
self.assertEqual(test_cgroup.path, "dummy_path")
self.assertEqual(test_cgroup.name, "test_extension")
def test_disable_should_reset_tracked_cgroups(self):
# Start tracking a couple of dummy cgroups
CGroupsTelemetry.track_cgroup(
CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service",
"cpu"))
CGroupsTelemetry.track_cgroup(
CGroup("dummy", "/sys/fs/cgroup/memory/system.slice/dummy.service",
"memory"))
CGroupConfiguratorSystemdTestCase._get_new_cgroup_configurator_instance(
).disable()
self.assertEquals(len(CGroupsTelemetry._tracked), 0)
def test_generate_extension_metrics_telemetry_dictionary(self, *args):
num_polls = 10
num_extensions = 1
cpu_percent_values = [random.randint(0, 100) for _ in range(num_polls)]
# only verifying calculations and not validity of the values.
memory_usage_values = [
random.randint(0, 8 * 1024**3) for _ in range(num_polls)
]
max_memory_usage_values = [
random.randint(0, 8 * 1024**3) for _ in range(num_polls)
]
# no need to initialize the CPU usage, since we mock get_cpu_usage() below
with patch(
"azurelinuxagent.common.cgroup.CpuCgroup.initialize_cpu_usage"
):
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create(
"dummy_cpu_path_{0}".format(i), "cpu",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create(
"dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
self.assertEqual(2 * num_extensions, len(CGroupsTelemetry._tracked))
with patch(
"azurelinuxagent.common.cgroup.MemoryCgroup.get_max_memory_usage"
) as patch_get_memory_max_usage:
with patch(
"azurelinuxagent.common.cgroup.MemoryCgroup.get_memory_usage"
) as patch_get_memory_usage:
with patch(
"azurelinuxagent.common.cgroup.CpuCgroup.get_cpu_usage"
) as patch_get_cpu_usage:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active"
) as patch_is_active:
for i in range(num_polls):
patch_is_active.return_value = True
patch_get_cpu_usage.return_value = cpu_percent_values[
i]
patch_get_memory_usage.return_value = memory_usage_values[
i] # example 200 MB
patch_get_memory_max_usage.return_value = max_memory_usage_values[
i] # example 450 MB
CGroupsTelemetry.poll_all_tracked()
def test_process_cgroup_metric_with_no_cpu_cgroup_mounted(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
with patch("azurelinuxagent.common.cgroup.MemoryCgroup._get_memory_max_usage") as patch_get_memory_max_usage:
with patch("azurelinuxagent.common.cgroup.MemoryCgroup._get_memory_usage") as patch_get_memory_usage:
with patch("azurelinuxagent.common.cgroup.CpuCgroup.collect") as patch_cpu_cgroup_collect:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active") as patch_is_active:
patch_is_active.return_value = True
patch_cpu_cgroup_collect.side_effect = Exception("File not found")
current_memory = 209715200
current_max_memory = 471859200
patch_get_memory_usage.return_value = current_memory # example 200 MB
patch_get_memory_max_usage.return_value = current_max_memory # example 450 MB
poll_count = 1
for data_count in range(poll_count, 10):
CGroupsTelemetry.poll_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
for cgroup_name, cgroup_metric in CGroupsTelemetry._cgroup_metrics.items():
current_memory_usage, max_memory_levels, current_cpu_usage = cgroup_metric.get_metrics()
self.assertEqual(len(current_memory_usage._data), data_count)
self.assertListEqual(current_memory_usage._data, [current_memory] * data_count)
self.assertEqual(len(max_memory_levels._data), data_count)
self.assertListEqual(max_memory_levels._data, [current_max_memory] * data_count)
self.assertEqual(len(current_cpu_usage._data), 0)
self.assertListEqual(current_cpu_usage._data, [])
CGroupsTelemetry.report_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
for cgroup_name, cgroup_metric in CGroupsTelemetry._cgroup_metrics.items():
current_memory_usage, max_memory_levels, current_cpu_usage = cgroup_metric.get_metrics()
self.assertEqual(len(current_memory_usage._data), 0)
self.assertEqual(len(max_memory_levels._data), 0)
self.assertEqual(len(current_cpu_usage._data), 0)
def test_is_active(self):
test_cgroup = CGroup.create(os.path.join(data_dir, "cgroups", "cpu_mount"), "cpu", "test_extension")
self.assertEqual(False, test_cgroup.is_active())
with open(os.path.join(data_dir, "cgroups", "cpu_mount", "tasks"), mode="wb") as tasks:
tasks.write(str(1000).encode())
self.assertEqual(True, test_cgroup.is_active())
test_cgroup = CGroup.create(os.path.join(data_dir, "cgroups", "memory_mount"), "memory", "test_extension")
self.assertEqual(False, test_cgroup.is_active())
with open(os.path.join(data_dir, "cgroups", "memory_mount", "tasks"), mode="wb") as tasks:
tasks.write(str(1000).encode())
self.assertEqual(True, test_cgroup.is_active())
def test_cgroup_is_tracked(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory", "dummy_extension_{0}".
format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
for i in range(num_extensions):
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
self.assertFalse(CGroupsTelemetry.is_tracked("not_present_cpu_dummy_path"))
self.assertFalse(CGroupsTelemetry.is_tracked("not_present_memory_dummy_path"))
def test_cgroup_tracking(self, *args):
num_extensions = 5
num_controllers = 2
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
for i in range(num_extensions):
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_cpu_path_{0}".format(i)))
self.assertTrue(CGroupsTelemetry.is_tracked("dummy_memory_path_{0}".format(i)))
self.assertEqual(num_extensions * num_controllers, len(CGroupsTelemetry._tracked))
def test_get_tracked_processes(self):
test_cgroup = CGroup.create(self.tmp_dir, "cpu", "test_extension")
self.assertListEqual(test_cgroup.get_tracked_processes(), [])
with open(os.path.join(self.tmp_dir, "cgroup.procs"),
mode="wb") as tasks:
tasks.write(str(1000).encode())
self.assertEqual(['1000'], test_cgroup.get_tracked_processes())
def test_send_extension_metrics_telemetry_for_unsupported_cgroup(
self, patch_periodic_warn, patch_add_metric, *args):
CGroupsTelemetry._tracked.append(
CGroup("cgroup_name", "/test/path", "io"))
PollResourceUsageOperation().run()
self.assertEqual(1, patch_periodic_warn.call_count)
self.assertEqual(
0, patch_add_metric.call_count) # No metrics should be sent.
def create_cgroup(controller):
path = FileSystemCgroupsApi._get_agent_cgroup_path(controller)
if not os.path.isdir(path):
FileSystemCgroupsApi._try_mkdir(path)
logger.info("Created cgroup {0}".format(path))
self._add_process_to_cgroup(pid, path)
cgroups.append(CGroup.create(path, controller, VM_AGENT_CGROUP_NAME))
def test_telemetry_polling_to_generate_transient_logs_index_error(self, mock_read_file, *args):
num_extensions = 1
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
# Generating a different kind of error (non-IOError) to check the logging.
# Trying to invoke IndexError during the getParameter call
mock_read_file.return_value = ''
with patch("azurelinuxagent.common.logger.periodic_warn") as patch_periodic_warn:
expected_call_count = 1 # called only once at start, and then gets removed from the tracked data.
for data_count in range(1, 10):
CGroupsTelemetry.poll_all_tracked()
self.assertEqual(expected_call_count, patch_periodic_warn.call_count)
def create_agent_cgroups(self):
try:
cgroup_unit = None
cgroup_paths = fileutil.read_file("/proc/self/cgroup")
for entry in cgroup_paths.splitlines():
fields = entry.split(':')
if fields[1] == "name=systemd":
cgroup_unit = fields[2].lstrip(os.path.sep)
cpu_cgroup_path = os.path.join(CGROUPS_FILE_SYSTEM_ROOT, 'cpu',
cgroup_unit)
memory_cgroup_path = os.path.join(CGROUPS_FILE_SYSTEM_ROOT,
'memory', cgroup_unit)
return [
CGroup.create(cpu_cgroup_path, 'cpu', cgroup_unit),
CGroup.create(memory_cgroup_path, 'memory', cgroup_unit)
]
except Exception as e:
raise CGroupsException(
"Failed to get paths of agent's cgroups. Error: {0}".format(
ustr(e)))
def test_telemetry_polling_to_not_generate_transient_logs_ioerror_file_not_found(self, mock_read_file,
patch_periodic_warn, *args):
num_extensions = 1
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
self.assertEqual(0, patch_periodic_warn.call_count)
# Not expecting logs present for io_error with errno=errno.ENOENT
io_error_2 = IOError()
io_error_2.errno = errno.ENOENT
mock_read_file.side_effect = io_error_2
poll_count = 1
for data_count in range(poll_count, 10):
CGroupsTelemetry.poll_all_tracked()
self.assertEqual(0, patch_periodic_warn.call_count)
def test_extension_temetry_not_sent_for_empty_perf_metrics(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
with patch("azurelinuxagent.common.cgroupstelemetry.CGroupsTelemetry._process_cgroup_metric") as \
patch_process_cgroup_metric:
with patch("azurelinuxagent.common.cgroup.CGroup.is_active") as patch_is_active:
patch_is_active.return_value = False
patch_process_cgroup_metric.return_value = {}
poll_count = 1
for data_count in range(poll_count, 10):
CGroupsTelemetry.poll_all_tracked()
collected_metrics = CGroupsTelemetry.report_all_tracked()
self.assertEqual(0, len(collected_metrics))
def test_process_cgroup_metric_with_incorrect_cgroups_mounted(self, *args):
num_extensions = 5
for i in range(num_extensions):
dummy_cpu_cgroup = CGroup.create("dummy_cpu_path_{0}".format(i), "cpu", "dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_cpu_cgroup)
dummy_memory_cgroup = CGroup.create("dummy_memory_path_{0}".format(i), "memory",
"dummy_extension_{0}".format(i))
CGroupsTelemetry.track_cgroup(dummy_memory_cgroup)
with patch("azurelinuxagent.common.cgroup.CpuCgroup.get_cpu_usage") as patch_get_cpu_usage:
with patch("azurelinuxagent.common.cgroup.MemoryCgroup.get_memory_usage") as patch_get_memory_usage:
patch_get_cpu_usage.side_effect = Exception("File not found")
patch_get_memory_usage.side_effect = Exception("File not found")
for data_count in range(1, 10):
CGroupsTelemetry.poll_all_tracked()
self.assertEqual(CGroupsTelemetry._cgroup_metrics.__len__(), num_extensions)
collected_metrics = {}
for name, cgroup_metrics in CGroupsTelemetry._cgroup_metrics.items():
collected_metrics[name] = CGroupsTelemetry._process_cgroup_metric(cgroup_metrics)
self.assertEqual(collected_metrics[name], {}) # empty
def test_send_extension_metrics_telemetry_for_unsupported_cgroup(
self, patch_periodic_warn, patch_add_metric, *args):
CGroupsTelemetry._tracked.append(
CGroup("cgroup_name", "/test/path", "io"))
monitor_handler = get_monitor_handler()
monitor_handler.init_protocols()
monitor_handler.last_cgroup_polling_telemetry = datetime.datetime.utcnow(
) - timedelta(hours=1)
monitor_handler.last_cgroup_report_telemetry = datetime.datetime.utcnow(
) - timedelta(hours=1)
monitor_handler.poll_telemetry_metrics()
self.assertEqual(1, patch_periodic_warn.call_count)
self.assertEqual(
0, patch_add_metric.call_count) # No metrics should be sent.
monitor_handler.stop()
def create_cgroup(controller):
try:
path = os.path.join(CGROUPS_FILE_SYSTEM_ROOT, controller,
VM_AGENT_CGROUP_NAME)
if not os.path.isdir(path):
FileSystemCgroupsApi._try_mkdir(path)
logger.info("Created cgroup {0}".format(path))
self._add_process_to_cgroup(pid, path)
cgroups.append(
CGroup.create(path, controller, VM_AGENT_CGROUP_NAME))
except Exception as e:
logger.warn(
'Cannot create "{0}" cgroup for the agent. Error: {1}'.
format(controller, ustr(e)))
def create_cgroup(controller):
cpu_cgroup_path = os.path.join(CGROUPS_FILE_SYSTEM_ROOT, controller, 'system.slice', slice_name)
cgroups.append(CGroup.create(cpu_cgroup_path, controller, extension_name))
def _create_extension_cgroup(self, controller, extension_name):
return CGroup.create(self._get_extension_cgroup_path(controller, extension_name), controller, extension_name)