def _TimeSeriesFromData(self, data, attr=None):
"""Build time series from StatsStore data."""
series = timeseries.Timeseries()
for value, timestamp in data:
if attr:
try:
series.Append(getattr(value, attr), timestamp)
except AttributeError:
raise ValueError("Can't find attribute %s in value %s." %
(attr, value))
else:
if hasattr(value, "sum") or hasattr(value, "count"):
raise ValueError(
"Can't treat complext type as simple value: %s" %
value)
series.Append(value, timestamp)
return series
def Handle(self, args, token=None):
start_time = args.start
end_time = args.end
if not end_time:
end_time = rdfvalue.RDFDatetime.Now()
if not start_time:
start_time = end_time - rdfvalue.Duration("30m")
fd = aff4.FACTORY.Create(args.client_id.ToClientURN().Add("stats"),
aff4_type=aff4_stats.ClientStats,
mode="r",
token=token,
age=(start_time, end_time))
stat_values = list(fd.GetValuesForAttribute(fd.Schema.STATS))
points = []
for stat_value in reversed(stat_values):
if args.metric == args.Metric.CPU_PERCENT:
points.extend((s.cpu_percent, s.timestamp)
for s in stat_value.cpu_samples)
elif args.metric == args.Metric.CPU_SYSTEM:
points.extend((s.system_cpu_time, s.timestamp)
for s in stat_value.cpu_samples)
elif args.metric == args.Metric.CPU_USER:
points.extend((s.user_cpu_time, s.timestamp)
for s in stat_value.cpu_samples)
elif args.metric == args.Metric.IO_READ_BYTES:
points.extend(
(s.read_bytes, s.timestamp) for s in stat_value.io_samples)
elif args.metric == args.Metric.IO_WRITE_BYTES:
points.extend((s.write_bytes, s.timestamp)
for s in stat_value.io_samples)
elif args.metric == args.Metric.IO_READ_OPS:
points.extend(
(s.read_count, s.timestamp) for s in stat_value.io_samples)
elif args.metric == args.Metric.IO_WRITE_OPS:
points.extend((s.write_count, s.timestamp)
for s in stat_value.io_samples)
elif args.metric == args.Metric.NETWORK_BYTES_RECEIVED:
points.append((stat_value.bytes_received, stat_value.age))
elif args.metric == args.Metric.NETWORK_BYTES_SENT:
points.append((stat_value.bytes_sent, stat_value.age))
elif args.metric == args.Metric.MEMORY_PERCENT:
points.append((stat_value.memory_percent, stat_value.age))
elif args.metric == args.Metric.MEMORY_RSS_SIZE:
points.append((stat_value.RSS_size, stat_value.age))
elif args.metric == args.Metric.MEMORY_VMS_SIZE:
points.append((stat_value.VMS_size, stat_value.age))
else:
raise ValueError("Unknown metric.")
# Points collected from "cpu_samples" and "io_samples" may not be correctly
# sorted in some cases (as overlaps between different stat_values are
# possible).
points.sort(key=lambda x: x[1])
ts = timeseries.Timeseries()
ts.MultiAppend(points)
if args.metric not in self.GAUGE_METRICS:
ts.MakeIncreasing()
if len(stat_values) > self.MAX_SAMPLES:
sampling_interval = rdfvalue.Duration.FromSeconds(
((end_time - start_time).seconds / self.MAX_SAMPLES) or 1)
if args.metric in self.GAUGE_METRICS:
mode = timeseries.NORMALIZE_MODE_GAUGE
else:
mode = timeseries.NORMALIZE_MODE_COUNTER
ts.Normalize(sampling_interval, start_time, end_time, mode=mode)
result = ApiGetClientLoadStatsResult()
for value, timestamp in ts.data:
dp = api_stats.ApiStatsStoreMetricDataPoint(timestamp=timestamp,
value=value)
result.data_points.append(dp)
return result
def makeSeries(self):
s = timeseries.Timeseries()
for i in range(1, 101):
s.Append(i, (i + 5) * 10000)
return s