def test_sets_buckets(self):
stats = MetricsBucketAggregator('myhost', interval=self.interval)
stats.submit_packets('my.set:10|s')
stats.submit_packets('my.set:20|s')
stats.submit_packets('my.set:20|s')
stats.submit_packets('my.set:30|s')
stats.submit_packets('my.set:30|s')
stats.submit_packets('my.set:30|s')
self.sleep_for_interval_length()
stats.submit_packets('my.set:40|s')
# Assert that it's treated normally.
self.sleep_for_interval_length()
metrics = stats.flush()
nt.assert_equal(len(metrics), 2)
m, m2 = metrics
nt.assert_equal(m['metric'], 'my.set')
nt.assert_equal(m['points'][0][1], 3)
nt.assert_equal(m2['metric'], 'my.set')
nt.assert_equal(m2['points'][0][1], 1)
# Assert there are no more sets
assert not stats.flush()
def test_histogram_flush_during_bucket(self):
ag_interval = 1
# The min is not enabled by default
stats = MetricsBucketAggregator(
'myhost',
interval=ag_interval,
histogram_aggregates=DEFAULT_HISTOGRAM_AGGREGATES + ['min'])
# Sample all numbers between 1-100 many times. This
# means our percentiles should be relatively close to themselves.
self.wait_for_bucket_boundary(ag_interval)
percentiles = range(100)
random.shuffle(percentiles) # in place
for i in percentiles:
for j in xrange(20):
for type_ in ['h', 'ms']:
m = 'my.p:%s|%s' % (i, type_)
stats.submit_packets(m)
time.sleep(
self.BUCKET_BOUNDARY_TOLERANCE
) # Make sure that we'll wait for the _next_ bucket boundary
self.wait_for_bucket_boundary(ag_interval)
percentiles = range(50)
random.shuffle(percentiles) # in place
for i in percentiles:
for j in xrange(20):
for type_ in ['h', 'ms']:
m = 'my.p:%s|%s' % (i, type_)
stats.submit_packets(m)
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 6)
p95, pavg, pcount, pmax, pmed, pmin = self.sort_metrics(metrics)
nt.assert_equal(p95['metric'], 'my.p.95percentile')
self.assert_almost_equal(p95['points'][0][1], 95, 10)
self.assert_almost_equal(pmax['points'][0][1], 99, 1)
self.assert_almost_equal(pmed['points'][0][1], 50, 2)
self.assert_almost_equal(pavg['points'][0][1], 50, 2)
self.assert_almost_equal(pmin['points'][0][1], 1, 1)
nt.assert_equal(pcount['points'][0][1], 4000) # 100 * 20 * 2
nt.assert_equals(p95['host'], 'myhost')
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 6)
p95_b, pavg_b, pcount_b, pmax_b, pmed_b, pmin_b = self.sort_metrics(
metrics)
nt.assert_equal(p95_b['metric'], 'my.p.95percentile')
self.assert_almost_equal(p95_b['points'][0][1], 47, 10)
self.assert_almost_equal(pmax_b['points'][0][1], 49, 1)
self.assert_almost_equal(pmed_b['points'][0][1], 25, 2)
self.assert_almost_equal(pavg_b['points'][0][1], 25, 2)
self.assert_almost_equal(pmin_b['points'][0][1], 1, 1)
nt.assert_equals(pcount_b['points'][0][1], 2000) # 100 * 20 * 2
# Ensure that histograms are reset.
metrics = self.sort_metrics(stats.flush())
assert not metrics
def test_sets_flush_during_bucket(self):
ag_interval = self.interval
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.wait_for_bucket_boundary(ag_interval)
stats.submit_packets('my.set:10|s')
stats.submit_packets('my.set:20|s')
stats.submit_packets('my.set:20|s')
stats.submit_packets('my.set:30|s')
stats.submit_packets('my.set:30|s')
stats.submit_packets('my.set:30|s')
self.sleep_for_interval_length(ag_interval)
stats.submit_packets('my.set:40|s')
# Assert that it's treated normally.
metrics = stats.flush()
nt.assert_equal(len(metrics), 1)
m = metrics[0]
nt.assert_equal(m['metric'], 'my.set')
nt.assert_equal(m['points'][0][1], 3)
self.sleep_for_interval_length(ag_interval)
metrics = stats.flush()
m = metrics[0]
nt.assert_equal(m['metric'], 'my.set')
nt.assert_equal(m['points'][0][1], 1)
# Assert there are no more sets
assert not stats.flush()
def test_gauge_flush_during_bucket(self):
#Tests returning data when flush is called in the middle of a time bucket that has data
ag_interval = self.interval
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.wait_for_bucket_boundary(ag_interval)
# Track some counters.
stats.submit_packets('my.first.gauge:1|g')
stats.submit_packets('my.first.gauge:5|g')
stats.submit_packets('my.second.gauge:1.5|g')
self.sleep_for_interval_length(ag_interval)
stats.submit_packets('my.second.gauge:9.5|g')
# Ensure that gauges roll up correctly.
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 2)
first, second = metrics
nt.assert_equals(first['metric'], 'my.first.gauge')
nt.assert_equals(first['points'][0][1], 5)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(second['metric'], 'my.second.gauge')
nt.assert_equals(second['points'][0][1], 1.5)
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 1)
nt.assert_equals(second['metric'], 'my.second.gauge')
nt.assert_equals(second['points'][0][1], 1.5)
def test_gauge_buckets(self):
# Tests calling returing data from 2 time buckets
ag_interval = self.interval
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.wait_for_bucket_boundary(ag_interval)
# Track some counters.
stats.submit_packets('my.first.gauge:1|g')
stats.submit_packets('my.first.gauge:5|g')
stats.submit_packets('my.second.gauge:1.5|g')
self.sleep_for_interval_length(ag_interval)
stats.submit_packets('my.second.gauge:9.5|g')
# Ensure that gauges roll up correctly.
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 3)
first, second, second_b = metrics
nt.assert_equals(first['metric'], 'my.first.gauge')
nt.assert_equals(first['points'][0][1], 5)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(second_b['metric'], 'my.second.gauge')
nt.assert_equals(second_b['points'][0][1], 9.5)
nt.assert_equals(second['metric'], 'my.second.gauge')
nt.assert_equals(second['points'][0][1], 1.5)
#check that they come back empty
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 0)
def test_counter_buckets(self):
ag_interval = 5
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.wait_for_bucket_boundary(ag_interval)
# Track some counters.
stats.submit_packets("my.first.counter:%s|c" % (1 * ag_interval))
stats.submit_packets("my.second.counter:%s|c" % (1 * ag_interval))
stats.submit_packets("my.third.counter:%s|c" % (3 * ag_interval))
time.sleep(ag_interval)
stats.submit_packets("my.first.counter:%s|c" % (5 * ag_interval))
# Want to get 2 different entries for my.first.counter in one set of metrics,
# so wait for the time bucket interval to pass
self.sleep_for_interval_length(ag_interval)
# Ensure they roll up nicely.
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 6)
first, first_b, second, second_b, third, third_b = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 1)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(first_b['metric'], 'my.first.counter')
nt.assert_equals(first_b['points'][0][1], 5)
nt.assert_equals(first_b['points'][0][0] - first['points'][0][0],
ag_interval)
nt.assert_equals(first['points'][0][0] % ag_interval, 0)
nt.assert_equals(first_b['points'][0][0] % ag_interval, 0)
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 1)
nt.assert_equals(second_b['metric'], 'my.second.counter')
nt.assert_equals(second_b['points'][0][1], 0)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 3)
nt.assert_equals(third_b['metric'], 'my.third.counter')
nt.assert_equals(third_b['points'][0][1], 0)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 0)
self.sleep_for_interval_length(ag_interval)
# Ensure that counters reset to zero.
metrics = self.sort_metrics(stats.flush())
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 0)
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 0)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 0)
def test_histogram_counter(self):
# Test whether histogram.count == increment
# same deal with a sample rate
ag_interval = self.interval
cnt = 100000
for run in [1, 2]:
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
for i in xrange(cnt):
if run == 2:
stats.submit_packets('test.counter:1|c|@0.5')
stats.submit_packets('test.hist:1|ms|@0.5')
else:
stats.submit_packets('test.counter:1|c')
stats.submit_packets('test.hist:1|ms')
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
assert len(metrics) > 0
#depending on timing, some runs may return the metric more that one bucket, meaning there may be
# more than one 'metric' for each of the counters
counter_count = 0
hist_count = 0
for num in [
m['points'][0][1] for m in metrics
if m['metric'] == 'test.counter'
]:
counter_count += num
for num in [
m['points'][0][1] for m in metrics
if m['metric'] == 'test.hist.count'
]:
hist_count += num
nt.assert_equal(counter_count, cnt * run)
nt.assert_equal(hist_count, cnt * run)
def test_tags(self):
stats = MetricsBucketAggregator('myhost', interval=self.interval)
stats.submit_packets('gauge:1|c')
stats.submit_packets('gauge:2|c|@1')
stats.submit_packets('gauge:4|c|#tag1,tag2')
stats.submit_packets(
'gauge:8|c|#tag2,tag1') # Should be the same as above
stats.submit_packets('gauge:16|c|#tag3,tag4')
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
assert len(metrics) == 3
first, second, third = metrics
nt.assert_equal(first['metric'], 'gauge')
nt.assert_equal(first['tags'], None)
nt.assert_equal(first['points'][0][1], 3)
nt.assert_equal(first['host'], 'myhost')
nt.assert_equal(second['metric'], 'gauge')
nt.assert_equal(second['tags'], ('tag1', 'tag2'))
nt.assert_equal(second['points'][0][1], 12)
nt.assert_equal(second['host'], 'myhost')
nt.assert_equal(third['metric'], 'gauge')
nt.assert_equal(third['tags'], ('tag3', 'tag4'))
nt.assert_equal(third['points'][0][1], 16)
nt.assert_equal(third['host'], 'myhost')
def test_sampled_counter(self):
# Submit a sampled counter.
stats = MetricsBucketAggregator('myhost', interval=self.interval)
stats.submit_packets('sampled.counter:1|c|@0.5')
self.sleep_for_interval_length()
metrics = stats.flush()
assert len(metrics) == 1
m = metrics[0]
assert m['metric'] == 'sampled.counter'
nt.assert_equal(m['points'][0][1], 2)
def test_scientific_notation(self):
ag_interval = 10
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
stats.submit_packets('test.scinot:9.512901e-05|g')
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
assert len(metrics) == 1
ts, val = metrics[0].get('points')[0]
nt.assert_almost_equal(val, 9.512901e-05)
def test_diagnostic_stats(self):
stats = MetricsBucketAggregator('myhost', interval=self.interval)
for i in xrange(10):
stats.submit_packets('metric:10|c')
stats.send_packet_count('a.stsstatsd.packet.count')
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(2, len(metrics))
first, second = metrics
nt.assert_equal(first['metric'], 'a.stsstatsd.packet.count')
nt.assert_equal(first['points'][0][1], 10)
def test_empty_counter(self):
ag_interval = self.interval
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
# Should be an empty list
nt.assert_equals(len(metrics), 0)
# Track some counters.
stats.submit_packets('my.first.counter:%s|c' % (1 * ag_interval))
# Call flush before the bucket_length has been exceeded
metrics = self.sort_metrics(stats.flush())
# Should be an empty list
nt.assert_equals(len(metrics), 0)
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
# Should now have the data
nt.assert_equals(len(metrics), 1)
nt.assert_equals(metrics[0]['metric'], 'my.first.counter')
nt.assert_equals(metrics[0]['points'][0][1], 1)
def test_string_sets(self):
stats = MetricsBucketAggregator('myhost', interval=self.interval)
stats.submit_packets('my.set:string|s')
stats.submit_packets('my.set:sets|s')
stats.submit_packets('my.set:sets|s')
stats.submit_packets('my.set:test|s')
stats.submit_packets('my.set:test|s')
stats.submit_packets('my.set:test|s')
# Assert that it's treated normally.
self.sleep_for_interval_length()
metrics = stats.flush()
nt.assert_equal(len(metrics), 1)
m = metrics[0]
nt.assert_equal(m['metric'], 'my.set')
nt.assert_equal(m['points'][0][1], 3)
# Assert there are no more sets
assert not stats.flush()
self.sleep_for_interval_length()
assert not stats.flush()
def test_batch_submission(self):
# Submit a sampled histogram.
stats = MetricsBucketAggregator('myhost', interval=self.interval)
metrics = ['counter:1|c', 'counter:1|c', 'gauge:1|g']
packet = "\n".join(metrics)
stats.submit_packets(packet)
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(2, len(metrics))
counter, gauge = metrics
assert counter['points'][0][1] == 2
assert gauge['points'][0][1] == 1
def test_gauge(self):
ag_interval = 2
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.wait_for_bucket_boundary(ag_interval)
# Track some counters.
stats.submit_packets('my.first.gauge:1|g')
stats.submit_packets('my.first.gauge:5|g')
stats.submit_packets('my.second.gauge:1.5|g')
# Ensure that gauges roll up correctly.
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 2)
first, second = metrics
nt.assert_equals(first['metric'], 'my.first.gauge')
nt.assert_equals(first['points'][0][1], 5)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(second['metric'], 'my.second.gauge')
nt.assert_equals(second['points'][0][1], 1.5)
# Ensure that old gauges get dropped due to old timestamps
stats.submit_metric('my.first.gauge', 5, 'g')
stats.submit_metric('my.first.gauge', 1, 'g', timestamp=1000000000)
stats.submit_metric('my.second.gauge', 20, 'g', timestamp=1000000000)
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 1)
first = metrics[0]
nt.assert_equals(first['metric'], 'my.first.gauge')
nt.assert_equals(first['points'][0][1], 5)
nt.assert_equals(first['host'], 'myhost')
def test_gauge_sample_rate(self):
ag_interval = self.interval
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
# Submit a sampled gauge metric.
stats.submit_packets('sampled.gauge:10|g|@0.1')
# Assert that it's treated normally.
self.sleep_for_interval_length(ag_interval)
metrics = stats.flush()
nt.assert_equal(len(metrics), 1)
m = metrics[0]
nt.assert_equal(m['metric'], 'sampled.gauge')
nt.assert_equal(m['points'][0][1], 10)
def test_counter(self):
ag_interval = 1.0
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
# Track some counters.
stats.submit_packets('my.first.counter:1|c')
stats.submit_packets('my.first.counter:5|c')
stats.submit_packets('my.second.counter:1|c')
stats.submit_packets('my.third.counter:3|c')
# Ensure they roll up nicely.
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
assert len(metrics) == 3
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 6)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 1)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 3)
self.sleep_for_interval_length(ag_interval)
# Ensure that counters reset to zero.
metrics = self.sort_metrics(stats.flush())
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 0)
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 0)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 0)
def test_sampled_histogram(self):
# Submit a sampled histogram.
# The min is not enabled by default
stats = MetricsBucketAggregator(
'myhost',
interval=self.interval,
histogram_aggregates=DEFAULT_HISTOGRAM_AGGREGATES + ['min'])
stats.submit_packets('sampled.hist:5|h|@0.5')
# Assert we scale up properly.
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
p95, pavg, pcount, pmax, pmed, pmin = self.sort_metrics(metrics)
nt.assert_equal(pcount['points'][0][1], 2)
for p in [p95, pavg, pmed, pmax, pmin]:
nt.assert_equal(p['points'][0][1], 5)
def test_histogram_normalization(self):
ag_interval = 10
# The min is not enabled by default
stats = MetricsBucketAggregator(
'myhost',
interval=ag_interval,
histogram_aggregates=DEFAULT_HISTOGRAM_AGGREGATES + ['min'])
for i in range(5):
stats.submit_packets('h1:1|h')
for i in range(20):
stats.submit_packets('h2:1|h')
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
_, _, h1count, _, _, _, _, _, h2count, _, _, _ = metrics
nt.assert_equal(h1count['points'][0][1], 0.5)
nt.assert_equal(h2count['points'][0][1], 2)
def test_counter_normalization(self):
ag_interval = 10
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
# Assert counters are normalized.
stats.submit_packets('int:1|c')
stats.submit_packets('int:4|c')
stats.submit_packets('int:15|c')
stats.submit_packets('float:5|c')
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
assert len(metrics) == 2
floatc, intc = metrics
nt.assert_equal(floatc['metric'], 'float')
nt.assert_equal(floatc['points'][0][1], 0.5)
nt.assert_equal(floatc['host'], 'myhost')
nt.assert_equal(intc['metric'], 'int')
nt.assert_equal(intc['points'][0][1], 2)
nt.assert_equal(intc['host'], 'myhost')
def test_recent_point_threshold(self):
ag_interval = 1
threshold = 100
# The min is not enabled by default
stats = MetricsBucketAggregator(
'myhost',
recent_point_threshold=threshold,
interval=ag_interval,
histogram_aggregates=DEFAULT_HISTOGRAM_AGGREGATES + ['min'])
timestamp_beyond_threshold = time.time() - threshold * 2
# Ensure that old gauges get dropped due to old timestamps
stats.submit_metric('my.first.gauge', 5, 'g')
stats.submit_metric('my.first.gauge',
1,
'g',
timestamp=timestamp_beyond_threshold)
stats.submit_metric('my.second.gauge',
20,
'g',
timestamp=timestamp_beyond_threshold)
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
assert len(metrics) == 1
first = metrics[0]
nt.assert_equals(first['metric'], 'my.first.gauge')
nt.assert_equals(first['points'][0][1], 5)
nt.assert_equals(first['host'], 'myhost')
timestamp_within_threshold = time.time() - threshold / 2
bucket_for_timestamp_within_threshold = timestamp_within_threshold - (
timestamp_within_threshold % ag_interval)
stats.submit_metric('my.1.gauge', 5, 'g')
stats.submit_metric('my.1.gauge',
1,
'g',
timestamp=timestamp_within_threshold)
stats.submit_metric('my.2.counter',
20,
'c',
timestamp=timestamp_within_threshold)
stats.submit_metric('my.3.set',
20,
's',
timestamp=timestamp_within_threshold)
stats.submit_metric('my.4.histogram',
20,
'h',
timestamp=timestamp_within_threshold)
self.sleep_for_interval_length(ag_interval)
flush_timestamp = time.time()
# The bucket timestamp is the beginning of the bucket that ended before we flushed
bucket_timestamp = flush_timestamp - (flush_timestamp %
ag_interval) - ag_interval
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 11)
first, first_b, second, second_b, third, h1, h2, h3, h4, h5, h6 = metrics
nt.assert_equals(first['metric'], 'my.1.gauge')
nt.assert_equals(first['points'][0][1], 1)
nt.assert_equals(first['host'], 'myhost')
self.assert_almost_equal(first['points'][0][0],
bucket_for_timestamp_within_threshold, 0.1)
nt.assert_equals(first_b['metric'], 'my.1.gauge')
nt.assert_equals(first_b['points'][0][1], 5)
self.assert_almost_equal(first_b['points'][0][0], bucket_timestamp,
0.1)
nt.assert_equals(second['metric'], 'my.2.counter')
nt.assert_equals(second['points'][0][1], 20)
self.assert_almost_equal(second['points'][0][0],
bucket_for_timestamp_within_threshold, 0.1)
nt.assert_equals(second_b['metric'], 'my.2.counter')
nt.assert_equals(second_b['points'][0][1], 0)
self.assert_almost_equal(second_b['points'][0][0], bucket_timestamp,
0.1)
nt.assert_equals(third['metric'], 'my.3.set')
nt.assert_equals(third['points'][0][1], 1)
self.assert_almost_equal(third['points'][0][0],
bucket_for_timestamp_within_threshold, 0.1)
nt.assert_equals(h1['metric'], 'my.4.histogram.95percentile')
nt.assert_equals(h1['points'][0][1], 20)
self.assert_almost_equal(h1['points'][0][0],
bucket_for_timestamp_within_threshold, 0.1)
nt.assert_equal(h1['points'][0][0], h2['points'][0][0])
nt.assert_equal(h1['points'][0][0], h3['points'][0][0])
nt.assert_equal(h1['points'][0][0], h4['points'][0][0])
nt.assert_equal(h1['points'][0][0], h5['points'][0][0])
def test_counter_flush_during_bucket(self):
ag_interval = 5
stats = MetricsBucketAggregator('myhost', interval=ag_interval)
self.wait_for_bucket_boundary(ag_interval)
time.sleep(0.5)
# Track some counters.
stats.submit_packets("my.first.counter:%s|c" % (1 * ag_interval))
stats.submit_packets("my.second.counter:%s|c" % (1 * ag_interval))
stats.submit_packets("my.third.counter:%s|c" % (3 * ag_interval))
time.sleep(ag_interval)
stats.submit_packets("my.first.counter:%s|c" % (5 * ag_interval))
# Want to get the date from the 2 buckets in 2 differnt calls, so don't wait for
# the bucket interval to pass
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 3)
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 1)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 1)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 3)
#Now wait for the bucket interval to pass, and get the other points
self.sleep_for_interval_length(ag_interval)
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 3)
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 5)
nt.assert_equals(first['host'], 'myhost')
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 0)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 0)
self.sleep_for_interval_length(ag_interval)
# Ensure that counters reset to zero.
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 3)
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 0)
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 0)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 0)
self.sleep_for_interval_length(ag_interval)
# Ensure that counters reset to zero.
metrics = self.sort_metrics(stats.flush())
nt.assert_equals(len(metrics), 3)
first, second, third = metrics
nt.assert_equals(first['metric'], 'my.first.counter')
nt.assert_equals(first['points'][0][1], 0)
nt.assert_equals(second['metric'], 'my.second.counter')
nt.assert_equals(second['points'][0][1], 0)
nt.assert_equals(third['metric'], 'my.third.counter')
nt.assert_equals(third['points'][0][1], 0)
def test_metrics_expiry(self):
# Ensure metrics eventually expire and stop submitting.
ag_interval = self.interval
expiry = ag_interval * 5 + 2
# The min is not enabled by default
stats = MetricsBucketAggregator(
'myhost',
interval=ag_interval,
expiry_seconds=expiry,
histogram_aggregates=DEFAULT_HISTOGRAM_AGGREGATES + ['min'])
stats.submit_packets('test.counter:123|c')
stats.submit_packets('test.gauge:55|g')
stats.submit_packets('test.set:44|s')
stats.submit_packets('test.histogram:11|h')
submit_time = time.time()
submit_bucket_timestamp = submit_time - (submit_time % ag_interval)
# Ensure points keep submitting
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 9)
nt.assert_equal(metrics[0]['metric'], 'test.counter')
nt.assert_equal(metrics[0]['points'][0][1], 123)
nt.assert_equal(metrics[0]['points'][0][0], submit_bucket_timestamp)
#flush without waiting - should get nothing
metrics = self.sort_metrics(stats.flush())
assert not metrics, str(metrics)
#Don't sumbit anything
submit_time = time.time()
bucket_timestamp = submit_time - (submit_time % ag_interval)
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 1)
nt.assert_equal(metrics[0]['metric'], 'test.counter')
nt.assert_equal(metrics[0]['points'][0][1], 0)
nt.assert_equal(metrics[0]['points'][0][0], bucket_timestamp)
stats.submit_packets('test.gauge:5|g')
self.sleep_for_interval_length()
time.sleep(0.3)
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 2)
nt.assert_equal(metrics[0]['metric'], 'test.counter')
nt.assert_equal(metrics[0]['points'][0][1], 0)
nt.assert_equal(metrics[1]['metric'], 'test.gauge')
nt.assert_equal(metrics[1]['points'][0][1], 5)
#flush without waiting - should get nothing
metrics = self.sort_metrics(stats.flush())
assert not metrics, str(metrics)
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 1)
nt.assert_equal(metrics[0]['metric'], 'test.counter')
nt.assert_equal(metrics[0]['points'][0][1], 0)
# Now sleep for longer than the expiry window and ensure
# no points are submitted
self.sleep_for_interval_length()
time.sleep(2)
m = stats.flush()
assert not m, str(m)
# If we submit again, we're all good.
stats.submit_packets('test.counter:123|c')
stats.submit_packets('test.gauge:55|g')
stats.submit_packets('test.set:44|s')
stats.submit_packets('test.histogram:11|h')
self.sleep_for_interval_length()
metrics = self.sort_metrics(stats.flush())
nt.assert_equal(len(metrics), 9)
nt.assert_equal(metrics[0]['metric'], 'test.counter')
nt.assert_equal(metrics[0]['points'][0][1], 123)