def test_detect_changes(self):
data = []
times = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]
values = [0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1]
for (t, v) in zip(times, values):
data.append(Datum(t, float(v)))
result = [(d.push_timestamp, d.state)
for d in detect_changes(data,
min_back_window=5,
max_back_window=5,
fore_window=5,
t_threshold=2)]
self.assertEqual(result, [(1, 'good'), (2, 'good'), (3, 'good'),
(4, 'good'), (5, 'good'), (6, 'good'),
(7, 'good'), (8, 'regression'), (9, 'good'),
(10, 'good')])
def check_json(self, filename, expected_timestamps):
"""Parse JSON produced by http://graphs.mozilla.org/api/test/runs"""
# Configuration for Analyzer
FORE_WINDOW = 12
MIN_BACK_WINDOW = 12
MAX_BACK_WINDOW = 24
THRESHOLD = 7
payload = SampleData.get_perf_data(os.path.join('graphs', filename))
runs = payload['test_runs']
data = []
for r in runs:
data.append(Datum(r[2], r[3], testrun_id=r[0],
revision_id=r[1][2]))
results = detect_changes(data, min_back_window=MIN_BACK_WINDOW,
max_back_window=MAX_BACK_WINDOW,
fore_window=FORE_WINDOW,
t_threshold=THRESHOLD)
regression_timestamps = [d.push_timestamp for d in results if
d.state == 'regression']
self.assertEqual(regression_timestamps, expected_timestamps)
def test_detect_changes(self):
data = []
times = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]
values = [0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1]
for (t, v) in zip(times, values):
data.append(Datum(t, float(v)))
result = [(d.push_timestamp, d.state) for d in
detect_changes(data, min_back_window=5, max_back_window=5,
fore_window=5, t_threshold=2)]
self.assertEqual(result, [
(1, 'good'),
(2, 'good'),
(3, 'good'),
(4, 'good'),
(5, 'good'),
(6, 'good'),
(7, 'good'),
(8, 'regression'),
(9, 'good'),
(10, 'good')])
def check_json(self, filename, expected_timestamps):
"""Parse JSON produced by http://graphs.mozilla.org/api/test/runs"""
# Configuration for Analyzer
FORE_WINDOW = 12
MIN_BACK_WINDOW = 12
MAX_BACK_WINDOW = 24
THRESHOLD = 7
payload = SampleData.get_perf_data(os.path.join('graphs', filename))
runs = payload['test_runs']
data = []
for r in runs:
data.append(Datum(r[2], r[3], testrun_id=r[0],
revision_id=r[1][2]))
results = detect_changes(data,
min_back_window=MIN_BACK_WINDOW,
max_back_window=MAX_BACK_WINDOW,
fore_window=FORE_WINDOW,
t_threshold=THRESHOLD)
regression_timestamps = [
d.push_timestamp for d in results if d.state == 'regression'
]
self.assertEqual(regression_timestamps, expected_timestamps)
def handle(self, *args, **options):
if options['server']:
server_params = urlparse(options['server'])
server_protocol = server_params.scheme
server_host = server_params.netloc
else:
server_protocol = settings.TREEHERDER_REQUEST_PROTOCOL
server_host = settings.TREEHERDER_REQUEST_HOST
if not options['project']:
raise CommandError("Must specify at least one project with "
"--project")
pc = PerfherderClient(protocol=server_protocol,
host=server_host)
option_collection_hash = pc.get_option_collection_hash()
# print csv header
print ','.join(["project", "platform", "signature", "series",
"testrun_id", "push_timestamp", "change",
"percent change", "t-value", "revision"])
for project in options['project']:
if options['signature']:
signatures = [options['signature']]
signature_data = pc.get_performance_signatures(
project, signatures=signatures,
interval=options['time_interval'])
else:
signature_data = pc.get_performance_signatures(
project, interval=options['time_interval'])
signatures = []
signatures_to_ignore = set()
# if doing everything, only handle summary series
for (signature, properties) in signature_data.iteritems():
signatures.append(signature)
if 'subtest_signatures' in properties:
# Don't alert on subtests which have a summary
signatures_to_ignore.update(properties['subtest_signatures'])
signatures = [signature for signature in signatures
if signature not in signatures_to_ignore]
for signature in signatures:
series = pc.get_performance_data(
project, signatures=signature,
interval=options['time_interval'])[signature]
series_properties = signature_data.get(signature)
data = []
for (result_set_id, timestamp, value) in zip(
series['result_set_id'], series['push_timestamp'],
series['value']):
data.append(Datum(timestamp, value, testrun_id=result_set_id))
for r in detect_changes(data):
if r.state == 'regression':
resultsets = pc.get_resultsets(project,
id=r.testrun_id)
if len(resultsets):
revision = resultsets[0]['revision']
else:
revision = ''
initial_value = r.historical_stats['avg']
new_value = r.forward_stats['avg']
if initial_value != 0:
pct_change = 100.0 * abs(new_value - initial_value) / float(initial_value)
else:
pct_change = 0.0
delta = (new_value - initial_value)
print ','.join(map(
lambda v: str(v),
[project, series_properties['machine_platform'],
signature, self._get_series_description(
option_collection_hash,
series_properties),
r.testrun_id, r.push_timestamp, delta,
pct_change, r.t, revision[0:12]]))
def generate_new_alerts_in_series(signature):
# get series data starting from either:
# (1) the last alert, if there is one
# (2) the alerts max age
# (use whichever is newer)
max_alert_age = (datetime.datetime.now() -
settings.PERFHERDER_ALERTS_MAX_AGE)
series = PerformanceDatum.objects.filter(signature=signature).filter(
push_timestamp__gte=max_alert_age).order_by('push_timestamp')
latest_alert_timestamp = PerformanceAlert.objects.filter(
series_signature=signature).select_related(
'summary__push__time').order_by(
'-summary__push__time').values_list(
'summary__push__time', flat=True)[:1]
if latest_alert_timestamp:
series = series.filter(
push_timestamp__gt=latest_alert_timestamp[0])
data = [Datum(int(time.mktime(d.push_timestamp.timetuple())), d.value, testrun_id=d.push_id) for d in series]
prev = None
min_back_window = signature.min_back_window
if min_back_window is None:
min_back_window = settings.PERFHERDER_ALERTS_MIN_BACK_WINDOW
max_back_window = signature.max_back_window
if max_back_window is None:
max_back_window = settings.PERFHERDER_ALERTS_MAX_BACK_WINDOW
fore_window = signature.fore_window
if fore_window is None:
fore_window = settings.PERFHERDER_ALERTS_FORE_WINDOW
alert_threshold = signature.alert_threshold
if alert_threshold is None:
alert_threshold = settings.PERFHERDER_REGRESSION_THRESHOLD
analyzed_series = detect_changes(data, min_back_window=min_back_window,
max_back_window=max_back_window,
fore_window=fore_window)
prev_testrun_id = None
with transaction.atomic():
for (prev, cur) in zip(analyzed_series, analyzed_series[1:]):
# we can have the same testrun id in a sequence if there are
# retriggers, so only set the prev_testrun_id if that isn't
# the case
if prev.testrun_id != cur.testrun_id:
prev_testrun_id = prev.testrun_id
if cur.state == 'regression' and prev_testrun_id:
prev_value = cur.historical_stats['avg']
new_value = cur.forward_stats['avg']
alert_properties = get_alert_properties(
prev_value, new_value, signature.lower_is_better)
if alert_properties.pct_change < alert_threshold:
# ignore regressions below the configured regression
# threshold
continue
summary, _ = PerformanceAlertSummary.objects.get_or_create(
repository=signature.repository,
framework=signature.framework,
push_id=cur.testrun_id,
prev_push_id=prev_testrun_id,
defaults={
'manually_created': False,
'last_updated': datetime.datetime.utcfromtimestamp(
cur.push_timestamp)
})
# django/mysql doesn't understand "inf", so just use some
# arbitrarily high value for that case
t_value = cur.t
if t_value == float('inf'):
t_value = 1000
PerformanceAlert.objects.update_or_create(
summary=summary,
series_signature=signature,
defaults={
'is_regression': alert_properties.is_regression,
'amount_pct': alert_properties.pct_change,
'amount_abs': alert_properties.delta,
'prev_value': prev_value,
'new_value': new_value,
't_value': t_value
})
def generate_new_alerts_in_series(signature):
# get series data starting from either:
# (1) the last alert, if there is one
# (2) the alerts max age
# (use whichever is newer)
max_alert_age = (datetime.datetime.now() -
settings.PERFHERDER_ALERTS_MAX_AGE)
series = PerformanceDatum.objects.filter(signature=signature).filter(
push_timestamp__gte=max_alert_age).order_by('push_timestamp')
latest_alert_timestamp = PerformanceAlert.objects.filter(
series_signature=signature).select_related(
'summary__push__time').order_by(
'-summary__push__time').values_list('summary__push__time',
flat=True)[:1]
if latest_alert_timestamp:
series = series.filter(push_timestamp__gt=latest_alert_timestamp[0])
data = [
Datum(int(time.mktime(d.push_timestamp.timetuple())),
d.value,
testrun_id=d.push_id) for d in series
]
prev = None
min_back_window = signature.min_back_window
if min_back_window is None:
min_back_window = settings.PERFHERDER_ALERTS_MIN_BACK_WINDOW
max_back_window = signature.max_back_window
if max_back_window is None:
max_back_window = settings.PERFHERDER_ALERTS_MAX_BACK_WINDOW
fore_window = signature.fore_window
if fore_window is None:
fore_window = settings.PERFHERDER_ALERTS_FORE_WINDOW
alert_threshold = signature.alert_threshold
if alert_threshold is None:
alert_threshold = settings.PERFHERDER_REGRESSION_THRESHOLD
analyzed_series = detect_changes(data,
min_back_window=min_back_window,
max_back_window=max_back_window,
fore_window=fore_window)
prev_testrun_id = None
with transaction.atomic():
for (prev, cur) in zip(analyzed_series, analyzed_series[1:]):
# we can have the same testrun id in a sequence if there are
# retriggers, so only set the prev_testrun_id if that isn't
# the case
if prev.testrun_id != cur.testrun_id:
prev_testrun_id = prev.testrun_id
if cur.state == 'regression' and prev_testrun_id:
prev_value = cur.historical_stats['avg']
new_value = cur.forward_stats['avg']
alert_properties = get_alert_properties(
prev_value, new_value, signature.lower_is_better)
if alert_properties.pct_change < alert_threshold:
# ignore regressions below the configured regression
# threshold
continue
summary, _ = PerformanceAlertSummary.objects.get_or_create(
repository=signature.repository,
framework=signature.framework,
push_id=cur.testrun_id,
prev_push_id=prev_testrun_id,
defaults={
'manually_created':
False,
'last_updated':
datetime.datetime.utcfromtimestamp(cur.push_timestamp)
})
# django/mysql doesn't understand "inf", so just use some
# arbitrarily high value for that case
t_value = cur.t
if t_value == float('inf'):
t_value = 1000
PerformanceAlert.objects.update_or_create(
summary=summary,
series_signature=signature,
defaults={
'is_regression': alert_properties.is_regression,
'amount_pct': alert_properties.pct_change,
'amount_abs': alert_properties.delta,
'prev_value': prev_value,
'new_value': new_value,
't_value': t_value
})
def handle(self, *args, **options):
if not options['project']:
raise CommandError("Must specify at least one project with "
"--project")
pc = PerfherderClient(server_url=options['server'])
option_collection_hash = pc.get_option_collection_hash()
# print csv header
print ','.join(["project", "platform", "signature", "series",
"testrun_id", "push_timestamp", "change",
"percent change", "t-value", "revision"])
for project in options['project']:
if options['signature']:
signatures = [options['signature']]
signature_data = pc.get_performance_signatures(
project, signatures=signatures,
interval=options['time_interval'])
else:
signature_data = pc.get_performance_signatures(
project, interval=options['time_interval'])
signatures = []
signatures_to_ignore = set()
# if doing everything, only handle summary series
for (signature, properties) in signature_data.iteritems():
signatures.append(signature)
if 'subtest_signatures' in properties:
# Don't alert on subtests which have a summary
signatures_to_ignore.update(properties['subtest_signatures'])
signatures = [signature for signature in signatures
if signature not in signatures_to_ignore]
for signature in signatures:
series = pc.get_performance_data(
project, signatures=signature,
interval=options['time_interval'])[signature]
series_properties = signature_data.get(signature)
data = []
for (result_set_id, timestamp, value) in zip(
series['result_set_id'], series['push_timestamp'],
series['value']):
data.append(Datum(timestamp, value, testrun_id=result_set_id))
for r in detect_changes(data):
if r.state == 'regression':
resultsets = pc.get_resultsets(project,
id=r.testrun_id)
if len(resultsets):
revision = resultsets[0]['revision']
else:
revision = ''
initial_value = r.historical_stats['avg']
new_value = r.forward_stats['avg']
if initial_value != 0:
pct_change = 100.0 * abs(new_value - initial_value) / float(initial_value)
else:
pct_change = 0.0
delta = (new_value - initial_value)
print ','.join(map(
lambda v: str(v),
[project, series_properties['machine_platform'],
signature, self._get_series_description(
option_collection_hash,
series_properties),
r.testrun_id, r.push_timestamp, delta,
pct_change, r.t, revision[0:12]]))
