def serialize(self, update):
"Generate serialized samples (id, dt, value) from a timestamped update dict."
for ts, data in update.items():
dt = datetime.fromtimestamp(ts, utc)
for name, item in data.items():
series = Series.get(self.measured_object, name, item["type"])
yield series.id, dt, item["value"]
def fetch_last(self, fetch_metrics):
"Return latest datetime and dict of field names and values."
latest, data = datetime.fromtimestamp(0, utc), {}
for series in Series.filter(self.measured_object, name__in=fetch_metrics):
point = Stats.latest(series.id)
data[series.name] = point.mean
latest = max(latest, point.dt)
return latest, data
def fetch_jobs(self,
metric,
begin,
end,
job,
max_points=float("inf"),
num_points=0):
"Return datetimes with dicts of field names and values."
result = collections.defaultdict(dict)
types = set()
begin = Stats[0].round(begin) # exclude points from a partial sample
end = Stats[0].round(end) # exclude points from a partial sample
series_ids = Series.filter(self.measured_object,
name__startswith="job_" +
metric).values("id")
series_ids = Stats[0].objects.filter(
id__in=series_ids, dt__gte=begin).values("id").distinct("id")
for series in Series.filter(self.measured_object, id__in=series_ids):
types.add(series.type)
for point in Stats.select(series.id,
begin,
end,
rate=True,
maxlen=max_points,
fixed=num_points):
result[point.dt][series.name.split("_", 3)[-1]] = point
assert types.issubset(Series.JOB_TYPES)
# translate job ids into metadata
metadata = dict((job_id, job_id) for points in result.values()
for job_id in points)
if job != "id":
for type in types: # there should generally be only one
metadata.update(scheduler.metadata(type, job, metadata))
for dt in result:
data = collections.defaultdict(lambda: Point.zero)
for job_id, point in result[dt].items():
data[metadata[job_id]] += point
result[dt] = dict((key, max(0.0, data[key].mean)) for key in data)
return dict(result)
def fetch(self, fetch_metrics, begin, end, max_points=float('inf'), num_points=0):
"Return datetimes with dicts of field names and values."
result = collections.defaultdict(dict)
types = set()
end = Stats[0].floor(end) # exclude points from a partial sample
for series in Series.filter(self.measured_object, name__in=fetch_metrics):
types.add(series.type)
minimum = 0.0 if series.type == 'Counter' else float('-inf')
for point in Stats.select(series.id, begin, end, rate=series.type in ('Counter', 'Derive'), maxlen=max_points, fixed=num_points):
result[point.dt][series.name] = max(minimum, point.mean)
# if absolute and derived values are mixed, the earliest value will be incomplete
if result and types > set(['Gauge']) and len(result[min(result)]) < len(fetch_metrics):
del result[min(result)]
return dict(result)
def names(self):
"names of all available data series"
return set(Series.filter(self.measured_object, type__in=Series.DATA_TYPES).values_list('name', flat=True))
def clear(self):
"Remove all associated series."
for series in Series.filter(self.measured_object):
series.delete()
Stats.delete(series.id)
