def test_query_structured_metrics(self):
mock_client, mock_job_result = self.setup_mock_client_result(
self.STRUCTURED_COUNTER_LIST)
dm = dataflow_metrics.DataflowMetrics(mock_client, mock_job_result)
dm._translate_step_name = types.MethodType(lambda self, x: 'split', dm)
query_result = dm.query()
expected_counters = [
MetricResult(
MetricKey(
'split',
MetricName('__main__.WordExtractingDoFn', 'word_lengths'),
), 109475, 109475),
]
self.assertEqual(query_result['counters'], expected_counters)
expected_distributions = [
MetricResult(
MetricKey(
'split',
MetricName('__main__.WordExtractingDoFn',
'word_length_dist'),
), DistributionResult(DistributionData(18, 2, 2, 16)),
DistributionResult(DistributionData(18, 2, 2, 16))),
]
self.assertEqual(query_result['distributions'], expected_distributions)
def test_direct_runner_metrics(self):
class MyDoFn(beam.DoFn):
def start_bundle(self):
count = Metrics.counter(self.__class__, 'bundles')
count.inc()
def finish_bundle(self):
count = Metrics.counter(self.__class__, 'finished_bundles')
count.inc()
def process(self, element):
gauge = Metrics.gauge(self.__class__, 'latest_element')
gauge.set(element)
count = Metrics.counter(self.__class__, 'elements')
count.inc()
distro = Metrics.distribution(self.__class__, 'element_dist')
distro.update(element)
return [element]
p = Pipeline(DirectRunner())
pcoll = (p | beam.Create([1, 2, 3, 4, 5])
| 'Do' >> beam.ParDo(MyDoFn()))
assert_that(pcoll, equal_to([1, 2, 3, 4, 5]))
result = p.run()
result.wait_until_finish()
metrics = result.metrics().query()
namespace = '{}.{}'.format(MyDoFn.__module__,
MyDoFn.__name__)
hc.assert_that(
metrics['counters'],
hc.contains_inanyorder(
MetricResult(
MetricKey('Do', MetricName(namespace, 'elements')),
5, 5),
MetricResult(
MetricKey('Do', MetricName(namespace, 'bundles')),
1, 1),
MetricResult(
MetricKey('Do', MetricName(namespace, 'finished_bundles')),
1, 1)))
hc.assert_that(
metrics['distributions'],
hc.contains_inanyorder(
MetricResult(
MetricKey('Do', MetricName(namespace, 'element_dist')),
DistributionResult(DistributionData(15, 5, 1, 5)),
DistributionResult(DistributionData(15, 5, 1, 5)))))
gauge_result = metrics['gauges'][0]
hc.assert_that(
gauge_result.key,
hc.equal_to(MetricKey('Do', MetricName(namespace, 'latest_element'))))
hc.assert_that(gauge_result.committed.value, hc.equal_to(5))
hc.assert_that(gauge_result.attempted.value, hc.equal_to(5))
def query(self, filter=None):
counters = [
MetricResult(MetricKey(k.step, k.metric), v.extract_committed(),
v.extract_latest_attempted())
for k, v in self._counters.items() if self.matches(filter, k)
]
distributions = [
MetricResult(MetricKey(k.step, k.metric), v.extract_committed(),
v.extract_latest_attempted())
for k, v in self._distributions.items() if self.matches(filter, k)
]
return {'counters': counters, 'distributions': distributions}
def test_direct_runner_metrics(self):
from apache_beam.metrics.metric import Metrics
class MyDoFn(beam.DoFn):
def start_bundle(self):
count = Metrics.counter(self.__class__, 'bundles')
count.inc()
def finish_bundle(self):
count = Metrics.counter(self.__class__, 'finished_bundles')
count.inc()
def process(self, element):
count = Metrics.counter(self.__class__, 'elements')
count.inc()
distro = Metrics.distribution(self.__class__, 'element_dist')
distro.update(element)
return [element]
runner = DirectRunner()
p = Pipeline(runner,
options=PipelineOptions(self.default_properties))
# pylint: disable=expression-not-assigned
(p | ptransform.Create([1, 2, 3, 4, 5])
| 'Do' >> beam.ParDo(MyDoFn()))
result = p.run()
result.wait_until_finish()
metrics = result.metrics().query()
namespace = '{}.{}'.format(MyDoFn.__module__,
MyDoFn.__name__)
hc.assert_that(
metrics['counters'],
hc.contains_inanyorder(
MetricResult(
MetricKey('Do', MetricName(namespace, 'elements')),
5, 5),
MetricResult(
MetricKey('Do', MetricName(namespace, 'bundles')),
1, 1),
MetricResult(
MetricKey('Do', MetricName(namespace, 'finished_bundles')),
1, 1)))
hc.assert_that(
metrics['distributions'],
hc.contains_inanyorder(
MetricResult(
MetricKey('Do', MetricName(namespace, 'element_dist')),
DistributionResult(DistributionData(15, 5, 1, 5)),
DistributionResult(DistributionData(15, 5, 1, 5)))))
def _combine(committed, attempted, filter):
all_keys = set(committed.keys()) | set(attempted.keys())
return [
MetricResult(key, committed.get(key), attempted.get(key))
for key in all_keys
if metric.MetricResults.matches(filter, key)
]
def test_query_counters(self):
mock_client, mock_job_result = self.setup_mock_client_result()
dm = dataflow_metrics.DataflowMetrics(mock_client, mock_job_result)
query_result = dm.query()
expected_counters = [
MetricResult(
MetricKey(
'split',
MetricName('__main__.WordExtractingDoFn', 'empty_lines')),
1080, 1080),
MetricResult(
MetricKey('longstepname/split',
MetricName('__main__.WordExtractingDoFn', 'words')),
26181, 26185),
]
self.assertEqual(
sorted(query_result['counters'], key=lambda x: x.key.metric.name),
sorted(expected_counters, key=lambda x: x.key.metric.name))
def query(self, filter=None):
counters = [
MetricResult(k, v, v) for k, v in self._counters.items()
if self.matches(filter, k)
]
distributions = [
MetricResult(k, v, v) for k, v in self._distributions.items()
if self.matches(filter, k)
]
gauges = [
MetricResult(k, v, v) for k, v in self._gauges.items()
if self.matches(filter, k)
]
return {
self.COUNTERS: counters,
self.DISTRIBUTIONS: distributions,
self.GAUGES: gauges
}
def test_query_counters(self):
mock_client, mock_job_result = self.setup_mock_client_result(
self.ONLY_COUNTERS_LIST)
dm = dataflow_metrics.DataflowMetrics(mock_client, mock_job_result)
dm._translate_step_name = types.MethodType(lambda self, x: 'split', dm)
query_result = dm.query()
expected_counters = [
MetricResult(
MetricKey(
'split',
MetricName('__main__.WordExtractingDoFn', 'empty_lines')),
1080, 1080),
MetricResult(
MetricKey('split',
MetricName('__main__.WordExtractingDoFn', 'words')),
26181, 26185),
]
self.assertEqual(
sorted(query_result['counters'], key=lambda x: x.key.metric.name),
sorted(expected_counters, key=lambda x: x.key.metric.name))
def query(self, filter=None):
counters = [MetricResult(MetricKey(k.step, k.metric),
v.extract_committed(),
v.extract_latest_attempted())
for k, v in self._counters.items()
if self.matches(filter, k)]
distributions = [MetricResult(MetricKey(k.step, k.metric),
v.extract_committed(),
v.extract_latest_attempted())
for k, v in self._distributions.items()
if self.matches(filter, k)]
gauges = [MetricResult(MetricKey(k.step, k.metric),
v.extract_committed(),
v.extract_latest_attempted())
for k, v in self._gauges.items()
if self.matches(filter, k)]
return {self.COUNTERS: counters,
self.DISTRIBUTIONS: distributions,
self.GAUGES: gauges}
def test_apply_physical_no_filter(self):
metrics = DirectMetrics()
metrics.update_physical(
object(),
MetricUpdates(
counters={
MetricKey('step1', self.name1): 5,
MetricKey('step1', self.name3): 8
}))
metrics.update_physical(
object(),
MetricUpdates(
counters={
MetricKey('step2', self.name1): 7,
MetricKey('step1', self.name3): 4
}))
results = metrics.query()
hc.assert_that(
results['counters'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1), 0, 5),
MetricResult(MetricKey('step1', self.name3), 0, 12),
MetricResult(MetricKey('step2', self.name1), 0, 7)
]))
metrics.commit_physical(object(), MetricUpdates())
results = metrics.query()
hc.assert_that(
results['counters'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1), 0, 5),
MetricResult(MetricKey('step1', self.name3), 0, 12),
MetricResult(MetricKey('step2', self.name1), 0, 7)
]))
def test_commit_logical_no_filter(self):
metrics = DirectMetrics()
metrics.commit_logical(
self.bundle1,
MetricUpdates(
counters={
MetricKey('step1', self.name1): 5,
MetricKey('step1', self.name2): 8
},
distributions={
MetricKey('step1', self.name1): DistributionData(8, 2, 3, 5)
}))
metrics.commit_logical(
self.bundle1,
MetricUpdates(
counters={
MetricKey('step2', self.name1): 7,
MetricKey('step1', self.name2): 4
},
distributions={
MetricKey('step1', self.name1): DistributionData(4, 1, 4, 4)
}))
results = metrics.query()
hc.assert_that(
results['counters'],
hc.contains_inanyorder(
*[
MetricResult(MetricKey('step1', self.name2), 12, 0),
MetricResult(MetricKey('step2', self.name1), 7, 0),
MetricResult(MetricKey('step1', self.name1), 5, 0)
]))
hc.assert_that(
results['distributions'],
hc.contains_inanyorder(
MetricResult(
MetricKey('step1', self.name1),
DistributionResult(DistributionData(12, 3, 3, 5)),
DistributionResult(DistributionData(0, 0, None, None)))))
def _populate_metrics(self, response, result, user_metrics=False):
"""Move metrics from response to results as MetricResults."""
if user_metrics:
metrics = [
metric for metric in response.metrics
if metric.name.origin == 'user'
]
else:
metrics = [
metric for metric in response.metrics
if metric.name.origin == 'dataflow/v1b3'
]
# Get the tentative/committed versions of every metric together.
metrics_by_name = defaultdict(lambda: {})
for metric in metrics:
if (metric.name.name.endswith('[MIN]')
or metric.name.name.endswith('[MAX]')
or metric.name.name.endswith('[MEAN]')
or metric.name.name.endswith('[COUNT]')):
# The Dataflow Service presents distribution metrics in two ways:
# One way is as a single distribution object with all its fields, and
# another way is as four different scalar metrics labeled as [MIN],
# [MAX], [COUNT], [MEAN].
# TODO(pabloem) remove these when distributions are not being broken up
# in the service.
# The second way is only useful for the UI, and should be ignored.
continue
is_tentative = [
prop for prop in metric.name.context.additionalProperties
if prop.key == 'tentative' and prop.value == 'true'
]
tentative_or_committed = 'tentative' if is_tentative else 'committed'
metric_key = self._get_metric_key(metric)
if metric_key is None:
continue
metrics_by_name[metric_key][tentative_or_committed] = metric
# Now we create the MetricResult elements.
for metric_key, metric in iteritems(metrics_by_name):
attempted = self._get_metric_value(metric['tentative'])
committed = self._get_metric_value(metric['committed'])
if attempted is None or committed is None:
continue
result.append(
MetricResult(metric_key,
attempted=attempted,
committed=committed))
def _populate_metric_results(self, response):
"""Take a list of metrics, and convert it to a list of MetricResult."""
user_metrics = [
metric for metric in response.metrics
if metric.name.origin == 'user'
]
# Get the tentative/committed versions of every metric together.
metrics_by_name = defaultdict(lambda: {})
for metric in user_metrics:
if (metric.name.name.endswith('(DIST)')
or metric.name.name.endswith('[MIN]')
or metric.name.name.endswith('[MAX]')
or metric.name.name.endswith('[MEAN]')
or metric.name.name.endswith('[COUNT]')):
warn(
'Distribution metrics will be ignored in the MetricsResult.query'
'method. You can see them in the Dataflow User Interface.')
# Distributions are not yet fully supported in this runner
continue
is_tentative = [
prop for prop in metric.name.context.additionalProperties
if prop.key == 'tentative' and prop.value == 'true'
]
tentative_or_committed = 'tentative' if is_tentative else 'committed'
metric_key = self._get_metric_key(metric)
metrics_by_name[metric_key][tentative_or_committed] = metric
# Now we create the MetricResult elements.
result = []
for metric_key, metric in metrics_by_name.iteritems():
if (metric['tentative'].scalar is None
or metric['committed'].scalar is None):
continue
attempted = metric['tentative'].scalar.integer_value
committed = metric['committed'].scalar.integer_value
result.append(
MetricResult(metric_key,
attempted=attempted,
committed=committed))
return result
def _create_metric_result(data_dict):
step = data_dict['step'] if 'step' in data_dict else ''
labels = data_dict['labels'] if 'labels' in data_dict else dict()
values = {}
for key in ['attempted', 'committed']:
if key in data_dict:
if 'counter' in data_dict[key]:
values[key] = data_dict[key]['counter']
elif 'distribution' in data_dict[key]:
distribution = data_dict[key]['distribution']
values[key] = DistributionResult(
DistributionData(
distribution['sum'],
distribution['count'],
distribution['min'],
distribution['max'],
))
attempted = values['attempted'] if 'attempted' in values else None
committed = values['committed'] if 'committed' in values else None
metric_name = MetricName(data_dict['namespace'], data_dict['name'])
metric_key = MetricKey(step, metric_name, labels)
return MetricResult(metric_key, committed, attempted)
def test_apply_physical_logical(self):
metrics = DirectMetrics()
dist_zero = DistributionData(0, 0, None, None)
metrics.update_physical(
object(),
MetricUpdates(counters={
MetricKey('step1', self.name1): 7,
MetricKey('step1', self.name2): 5,
MetricKey('step2', self.name1): 1
},
distributions={
MetricKey('step1', self.name1):
DistributionData(3, 1, 3, 3),
MetricKey('step2', self.name3):
DistributionData(8, 2, 4, 4)
}))
results = metrics.query()
hc.assert_that(
results['counters'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1), 0, 7),
MetricResult(MetricKey('step1', self.name2), 0, 5),
MetricResult(MetricKey('step2', self.name1), 0, 1)
]))
hc.assert_that(
results['distributions'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1),
DistributionResult(dist_zero),
DistributionResult(DistributionData(3, 1, 3, 3))),
MetricResult(MetricKey('step2', self.name3),
DistributionResult(dist_zero),
DistributionResult(DistributionData(8, 2, 4, 4)))
]))
metrics.commit_physical(
object(),
MetricUpdates(counters={
MetricKey('step1', self.name1): -3,
MetricKey('step2', self.name1): -5
},
distributions={
MetricKey('step1', self.name1):
DistributionData(8, 4, 1, 5),
MetricKey('step2', self.name2):
DistributionData(8, 8, 1, 1)
}))
results = metrics.query()
hc.assert_that(
results['counters'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1), 0, 4),
MetricResult(MetricKey('step1', self.name2), 0, 5),
MetricResult(MetricKey('step2', self.name1), 0, -4)
]))
hc.assert_that(
results['distributions'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1),
DistributionResult(dist_zero),
DistributionResult(DistributionData(11, 5, 1,
5))),
MetricResult(MetricKey('step2', self.name3),
DistributionResult(dist_zero),
DistributionResult(DistributionData(8, 2, 4, 4))),
MetricResult(MetricKey('step2', self.name2),
DistributionResult(dist_zero),
DistributionResult(DistributionData(8, 8, 1, 1)))
]))
metrics.commit_logical(
object(),
MetricUpdates(counters={
MetricKey('step1', self.name1): 3,
MetricKey('step1', self.name2): 5,
MetricKey('step2', self.name1): -3
},
distributions={
MetricKey('step1', self.name1):
DistributionData(11, 5, 1, 5),
MetricKey('step2', self.name2):
DistributionData(8, 8, 1, 1),
MetricKey('step2', self.name3):
DistributionData(4, 1, 4, 4)
}))
results = metrics.query()
hc.assert_that(
results['counters'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1), 3, 4),
MetricResult(MetricKey('step1', self.name2), 5, 5),
MetricResult(MetricKey('step2', self.name1), -3, -4)
]))
hc.assert_that(
results['distributions'],
hc.contains_inanyorder(*[
MetricResult(MetricKey('step1', self.name1),
DistributionResult(DistributionData(11, 5, 1, 5)),
DistributionResult(DistributionData(11, 5, 1,
5))),
MetricResult(MetricKey('step2', self.name3),
DistributionResult(DistributionData(4, 1, 4, 4)),
DistributionResult(DistributionData(8, 2, 4, 4))),
MetricResult(MetricKey('step2', self.name2),
DistributionResult(DistributionData(8, 8, 1, 1)),
DistributionResult(DistributionData(8, 8, 1, 1)))
]))
