def get_failures(self, node_id, threshold_secs=90, upper_failure_threshold=0.2,
after=_AFTER_DEFAULT,
before=_BEFORE_DEFAULT):
"""Find websites which have experienced two or more consecutive failures
"""
probes_count = self.get_probes_count(node_id)
c = self.db.cursor()
c.execute('select website, end_time from metrics where reason != 0 and end_time >= %s and end_time <= %s'
'and node_id = %s order by end_time',
(datetime_to_mysql_date(after), datetime_to_mysql_date(before), node_id))
failed_websites = set()
website2failure_time = defaultdict(list)
for _ in range(c.rowcount):
website, end_time = c.fetchone()
if not website2failure_time[website]:
website2failure_time[website].append(end_time)
else:
if end_time <= website2failure_time[website][-1] + datetime.timedelta(seconds=threshold_secs):
failed_websites.add(website)
website2failure_time[website].append(end_time)
c.close()
for k, v in list(website2failure_time.iteritems()):
# I don't think it makes sense to take seriously websites which fail
# more than certain rate
if 1.0 * len(v) / probes_count > upper_failure_threshold:
failed_websites.remove(k)
return {w: website2failure_time[w] for w in failed_websites}
def _metric_to_mysql_tuple(self, m):
return (m.node_id,
m.website,
m.state,
datetime_to_mysql_date(m.start),
datetime_to_mysql_date(m.end),
(m.end - m.start).total_seconds(), # duration in seconds
0 if m.http_code is None else m.http_code,
# normalized data
datetime_to_mysql_date(m.end_1min),
datetime_to_mysql_date(m.end_5min),
0 if m.state == Metric.STATE_OK else 1
)
def plot_duration_and_failure(self, website,
after=_AFTER_DEFAULT,
before=_BEFORE_DEFAULT):
c = self.db.cursor()
c.execute('select sum(failure), avg(duration), end_time_1min from metrics '
'where website=%s '
'and end_time >= %s and end_time <= %s '
'group by end_time_1min '
'having count(end_time_1min) >= 2',
(website, datetime_to_mysql_date(after), datetime_to_mysql_date(before)))
result = {'time': [], 'sum(failure)': [], 'avg(duration)': []}
for _ in range(c.rowcount):
failure, duration, time = c.fetchone()
result['time'].append(time)
result['sum(failure)'].append(int(failure))
result['avg(duration)'].append(float(duration))
c.close()
df = pd.DataFrame(result)
return df.plot(x='time', title=website, secondary_y='sum(failure)', figsize=(16, 12))
