/usr/lib/python2.7/dist-packages/gnocchi/carbonara.py is in python-gnocchi 2.0.2-4.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 | # -*- encoding: utf-8 -*-
#
# Copyright © 2016 Red Hat, Inc.
# Copyright © 2014-2015 eNovance
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""Time series data manipulation, better with pancetta."""
import datetime
import functools
import logging
import numbers
import operator
import re
import iso8601
import lz4
import msgpack
import pandas
import six
from gnocchi import utils
LOG = logging.getLogger(__name__)
class NoDeloreanAvailable(Exception):
"""Error raised when trying to insert a value that is too old."""
def __init__(self, first_timestamp, bad_timestamp):
self.first_timestamp = first_timestamp
self.bad_timestamp = bad_timestamp
super(NoDeloreanAvailable, self).__init__(
"%s is before %s" % (bad_timestamp, first_timestamp))
class UnAggregableTimeseries(Exception):
"""Error raised when timeseries cannot be aggregated."""
def __init__(self, reason):
self.reason = reason
super(UnAggregableTimeseries, self).__init__(reason)
class UnknownAggregationMethod(Exception):
"""Error raised when the aggregation method is unknown."""
def __init__(self, agg):
self.aggregation_method = agg
super(UnknownAggregationMethod, self).__init__(
"Unknown aggregation method `%s'" % agg)
class SerializableMixin(object):
@classmethod
def unserialize(cls, data):
return cls.from_dict(msgpack.loads(data, encoding='utf-8'))
def serialize(self):
return msgpack.dumps(self.to_dict())
class TimeSerie(SerializableMixin):
"""A representation of series of a timestamp with a value.
Duplicate timestamps are not allowed and will be filtered to use the
last in the group when the TimeSerie is created or extended.
"""
def __init__(self, ts=None):
if ts is None:
ts = pandas.Series()
self.ts = self.clean_ts(ts)
@staticmethod
def clean_ts(ts):
if ts.index.has_duplicates:
ts = ts[~ts.index.duplicated(keep='last')]
if not ts.index.is_monotonic:
ts = ts.sort_index()
return ts
@classmethod
def from_data(cls, timestamps=None, values=None):
return cls(pandas.Series(values, timestamps))
@classmethod
def from_tuples(cls, timestamps_values):
return cls.from_data(*zip(*timestamps_values))
def __eq__(self, other):
return (isinstance(other, TimeSerie)
and self.ts.all() == other.ts.all())
def __getitem__(self, key):
return self.ts[key]
def set_values(self, values):
t = pandas.Series(*reversed(list(zip(*values))))
self.ts = self.clean_ts(t).combine_first(self.ts)
def __len__(self):
return len(self.ts)
@staticmethod
def _timestamps_and_values_from_dict(values):
v = tuple(
zip(*dict(
(pandas.Timestamp(k), v)
for k, v in six.iteritems(values)).items()))
if v:
return v
return (), ()
@classmethod
def from_dict(cls, d):
"""Build a time series from a dict.
The dict format must be datetime as key and values as values.
:param d: The dict.
:returns: A TimeSerie object
"""
return cls.from_data(
*cls._timestamps_and_values_from_dict(d['values']))
def to_dict(self):
return {
'values': dict((timestamp.value, float(v))
for timestamp, v
in six.iteritems(self.ts.dropna())),
}
@staticmethod
def _serialize_time_period(value):
if value:
return value.nanos / 10e8
@staticmethod
def _round_timestamp(ts, freq):
return pandas.Timestamp(
(pandas.Timestamp(ts).value // freq) * freq)
@staticmethod
def _to_offset(value):
if isinstance(value, numbers.Real):
return pandas.tseries.offsets.Nano(value * 10e8)
return pandas.tseries.frequencies.to_offset(value)
@property
def first(self):
try:
return self.ts.index[0]
except IndexError:
return
@property
def last(self):
try:
return self.ts.index[-1]
except IndexError:
return
class BoundTimeSerie(TimeSerie):
def __init__(self, ts=None, block_size=None, back_window=0):
"""A time serie that is limited in size.
Used to represent the full-resolution buffer of incoming raw
datapoints associated with a metric.
The maximum size of this time serie is expressed in a number of block
size, called the back window.
When the timeserie is truncated, a whole block is removed.
You cannot set a value using a timestamp that is prior to the last
timestamp minus this number of blocks. By default, a back window of 0
does not allow you to go back in time prior to the current block being
used.
"""
super(BoundTimeSerie, self).__init__(ts)
self.block_size = self._to_offset(block_size)
self.back_window = back_window
self._truncate()
@classmethod
def from_data(cls, timestamps=None, values=None,
block_size=None, back_window=0):
return cls(pandas.Series(values, timestamps),
block_size=block_size, back_window=back_window)
def __eq__(self, other):
return (isinstance(other, BoundTimeSerie)
and super(BoundTimeSerie, self).__eq__(other)
and self.block_size == other.block_size
and self.back_window == other.back_window)
def set_values(self, values, before_truncate_callback=None,
ignore_too_old_timestamps=False):
if self.block_size is not None and not self.ts.empty:
values = sorted(values, key=operator.itemgetter(0))
first_block_timestamp = self._first_block_timestamp()
if ignore_too_old_timestamps:
for index, (timestamp, value) in enumerate(values):
if timestamp >= first_block_timestamp:
values = values[index:]
break
else:
values = []
else:
# Check that the smallest timestamp does not go too much back
# in time.
smallest_timestamp = values[0][0]
if smallest_timestamp < first_block_timestamp:
raise NoDeloreanAvailable(first_block_timestamp,
smallest_timestamp)
super(BoundTimeSerie, self).set_values(values)
if before_truncate_callback:
before_truncate_callback(self)
self._truncate()
@classmethod
def from_dict(cls, d):
"""Build a time series from a dict.
The dict format must be datetime as key and values as values.
:param d: The dict.
:returns: A TimeSerie object
"""
timestamps, values = cls._timestamps_and_values_from_dict(d['values'])
return cls.from_data(timestamps, values,
block_size=d.get('block_size'),
back_window=d.get('back_window'))
def to_dict(self):
basic = super(BoundTimeSerie, self).to_dict()
basic.update({
'block_size': self._serialize_time_period(self.block_size),
'back_window': self.back_window,
})
return basic
def _first_block_timestamp(self):
rounded = self._round_timestamp(self.ts.index[-1],
self.block_size.delta.value)
return rounded - (self.block_size * self.back_window)
def _truncate(self):
"""Truncate the timeserie."""
if self.block_size is not None and not self.ts.empty:
# Change that to remove the amount of block needed to have
# the size <= max_size. A block is a number of "seconds" (a
# timespan)
self.ts = self.ts[self._first_block_timestamp():]
class AggregatedTimeSerie(TimeSerie):
_AGG_METHOD_PCT_RE = re.compile(r"([1-9][0-9]?)pct")
POINTS_PER_SPLIT = 14400
def __init__(self, sampling, ts=None, max_size=None,
aggregation_method='mean'):
"""A time serie that is downsampled.
Used to represent the downsampled timeserie for a single
granularity/aggregation-function pair stored for a metric.
"""
super(AggregatedTimeSerie, self).__init__(ts)
m = self._AGG_METHOD_PCT_RE.match(aggregation_method)
if m:
self.q = float(m.group(1)) / 100
self.aggregation_method_func_name = 'quantile'
else:
if not hasattr(pandas.core.groupby.SeriesGroupBy,
aggregation_method):
raise UnknownAggregationMethod(aggregation_method)
self.aggregation_method_func_name = aggregation_method
self.sampling = self._to_offset(sampling).nanos / 10e8
self.max_size = max_size
self.aggregation_method = aggregation_method
@classmethod
def from_data(cls, sampling, timestamps=None, values=None,
max_size=None, aggregation_method='mean'):
return cls(ts=pandas.Series(values, timestamps),
max_size=max_size, sampling=sampling,
aggregation_method=aggregation_method)
@classmethod
def get_split_key_datetime(cls, timestamp, sampling):
return cls._round_timestamp(
timestamp, freq=sampling * cls.POINTS_PER_SPLIT * 10e8)
@staticmethod
def _split_key_to_string(timestamp):
ts = timestamp.to_datetime()
if ts.tzinfo is None:
ts = ts.replace(tzinfo=iso8601.iso8601.UTC)
return str(utils.datetime_to_unix(ts))
@classmethod
def get_split_key(cls, timestamp, sampling):
return cls._split_key_to_string(
cls.get_split_key_datetime(timestamp, sampling))
def split(self):
groupby = self.ts.groupby(functools.partial(
self.get_split_key_datetime, sampling=self.sampling))
keys = sorted(groupby.groups.keys())
for i, ts in enumerate(keys):
if i + 1 == len(keys):
yield self._split_key_to_string(ts), TimeSerie(self.ts[ts:])
elif i + 1 < len(keys):
t = self.ts[ts:keys[i + 1]]
del t[t.index[-1]]
yield self._split_key_to_string(ts), TimeSerie(t)
@classmethod
def from_timeseries(cls, timeseries, sampling, max_size=None,
aggregation_method='mean'):
ts = pandas.Series()
for t in timeseries:
ts = ts.combine_first(t.ts)
return cls(ts=ts, sampling=sampling, max_size=max_size,
aggregation_method=aggregation_method)
def __eq__(self, other):
return (isinstance(other, AggregatedTimeSerie)
and super(AggregatedTimeSerie, self).__eq__(other)
and self.max_size == other.max_size
and self.sampling == other.sampling
and self.aggregation_method == other.aggregation_method)
def __repr__(self):
return "<%s 0x%x sampling=%fs max_size=%s agg_method=%s>" % (
self.__class__.__name__,
id(self),
self.sampling,
self.max_size,
self.aggregation_method,
)
@classmethod
def from_dict(cls, d):
"""Build a time series from a dict.
The dict format must be datetime as key and values as values.
:param d: The dict.
:returns: A TimeSerie object
"""
sampling = d.get('sampling')
if 'first_timestamp' in d:
prev_timestamp = pandas.Timestamp(d.get('first_timestamp') * 10e8)
timestamps = []
for delta in d.get('timestamps'):
prev_timestamp = datetime.timedelta(
seconds=delta * sampling) + prev_timestamp
timestamps.append(prev_timestamp)
else:
# migrate from v1.3, remove with TimeSerieArchive
timestamps, d['values'] = (
cls._timestamps_and_values_from_dict(d['values']))
return cls.from_data(
timestamps=timestamps,
values=d.get('values'),
max_size=d.get('max_size'),
sampling=sampling,
aggregation_method=d.get('aggregation_method', 'mean'))
def to_dict(self):
if self.ts.empty:
timestamps = []
values = []
first_timestamp = 0
else:
first_timestamp = float(
self.get_split_key(self.ts.index[0], self.sampling))
timestamps = []
prev_timestamp = pandas.Timestamp(
first_timestamp * 10e8).to_pydatetime()
# Use double delta encoding for timestamps
for i in self.ts.index:
# Convert to pydatetime because it's faster to compute than
# Pandas' objects
asdt = i.to_pydatetime()
timestamps.append(
int((asdt - prev_timestamp).total_seconds()
/ self.sampling))
prev_timestamp = asdt
values = self.ts.values.tolist()
return {
'first_timestamp': first_timestamp,
'aggregation_method': self.aggregation_method,
'max_size': self.max_size,
'sampling': self.sampling,
'timestamps': timestamps,
'values': values,
}
@classmethod
def unserialize(cls, data):
return cls.from_dict(msgpack.loads(lz4.loads(data), encoding='utf-8'))
def serialize(self):
return lz4.dumps(msgpack.dumps(self.to_dict()))
def _truncate(self):
"""Truncate the timeserie."""
if self.max_size is not None:
# Remove empty points if any that could be added by aggregation
self.ts = self.ts.dropna()[-self.max_size:]
def _resample(self, after):
# Group by the sampling, and then apply the aggregation method on
# the points after `after'
groupedby = self.ts[after:].groupby(
functools.partial(self._round_timestamp,
freq=self.sampling * 10e8))
agg_func = getattr(groupedby, self.aggregation_method_func_name)
if self.aggregation_method_func_name == 'quantile':
aggregated = agg_func(self.q)
else:
aggregated = agg_func()
# Now combine the result with the rest of the point – everything
# that is before `after'
self.ts = aggregated.combine_first(self.ts[:after][:-1])
def fetch(self, from_timestamp=None, to_timestamp=None):
"""Fetch aggregated time value.
Returns a sorted list of tuples (timestamp, granularity, value).
"""
# Round timestamp to our granularity so we're sure that if e.g. 17:02
# is requested and we have points for 17:00 and 17:05 in a 5min
# granularity, we do return the 17:00 point and not nothing
if from_timestamp is None:
from_ = None
else:
from_ = self._round_timestamp(from_timestamp, self.sampling * 10e8)
points = self[from_:to_timestamp]
try:
# Do not include stop timestamp
del points[to_timestamp]
except KeyError:
pass
return [(timestamp, self.sampling, value)
for timestamp, value
in six.iteritems(points)]
def update(self, ts):
if ts.ts.empty:
return
ts.ts = self.clean_ts(ts.ts)
index = ts.ts.index
first_timestamp = index[0]
last_timestamp = index[-1]
# Build a new time serie excluding all data points in the range of the
# timeserie passed as argument
new_ts = self.ts.drop(self.ts[first_timestamp:last_timestamp].index)
# Build a new timeserie where we replaced the timestamp range covered
# by the timeserie passed as argument
self.ts = ts.ts.combine_first(new_ts)
# Resample starting from the first timestamp we received
# TODO(jd) So this only works correctly because we expect that we are
# not going to replace a range in the middle of our timeserie. So we re
# resample EVERYTHING FROM first timestamp. We should rather resample
# from first timestamp AND TO LAST TIMESTAMP!
self._resample(first_timestamp)
self._truncate()
@staticmethod
def aggregated(timeseries, from_timestamp=None, to_timestamp=None,
aggregation='mean', needed_percent_of_overlap=100.0):
index = ['timestamp', 'granularity']
columns = ['timestamp', 'granularity', 'value']
dataframes = []
if not timeseries:
return []
for timeserie in timeseries:
timeserie_raw = timeserie.fetch(from_timestamp, to_timestamp)
if timeserie_raw:
dataframe = pandas.DataFrame(timeserie_raw, columns=columns)
dataframe = dataframe.set_index(index)
dataframes.append(dataframe)
if not dataframes:
return []
number_of_distinct_datasource = len(timeseries) / len(
set(ts.sampling for ts in timeseries)
)
grouped = pandas.concat(dataframes).groupby(level=index)
left_boundary_ts = None
right_boundary_ts = None
maybe_next_timestamp_is_left_boundary = False
left_holes = 0
right_holes = 0
holes = 0
for (timestamp, __), group in grouped:
if group.count()['value'] != number_of_distinct_datasource:
maybe_next_timestamp_is_left_boundary = True
if left_boundary_ts is not None:
right_holes += 1
else:
left_holes += 1
elif maybe_next_timestamp_is_left_boundary:
left_boundary_ts = timestamp
maybe_next_timestamp_is_left_boundary = False
else:
right_boundary_ts = timestamp
holes += right_holes
right_holes = 0
if to_timestamp is not None:
holes += left_holes
if from_timestamp is not None:
holes += right_holes
if to_timestamp is not None or from_timestamp is not None:
maximum = len(grouped)
percent_of_overlap = (float(maximum - holes) * 100.0 /
float(maximum))
if percent_of_overlap < needed_percent_of_overlap:
raise UnAggregableTimeseries(
'Less than %f%% of datapoints overlap in this '
'timespan (%.2f%%)' % (needed_percent_of_overlap,
percent_of_overlap))
if (needed_percent_of_overlap > 0 and
(right_boundary_ts == left_boundary_ts or
(right_boundary_ts is None
and maybe_next_timestamp_is_left_boundary))):
LOG.debug("We didn't find points that overlap in those "
"timeseries. "
"right_boundary_ts=%(right_boundary_ts)s, "
"left_boundary_ts=%(left_boundary_ts)s, "
"groups=%(groups)s" % {
'right_boundary_ts': right_boundary_ts,
'left_boundary_ts': left_boundary_ts,
'groups': list(grouped)
})
raise UnAggregableTimeseries('No overlap')
# NOTE(sileht): this call the aggregation method on already
# aggregated values, for some kind of aggregation this can
# result can looks weird, but this is the best we can do
# because we don't have anymore the raw datapoints in those case.
# FIXME(sileht): so should we bailout is case of stddev, percentile
# and median?
agg_timeserie = getattr(grouped, aggregation)()
agg_timeserie = agg_timeserie.dropna().reset_index()
if from_timestamp is None and left_boundary_ts:
agg_timeserie = agg_timeserie[
agg_timeserie['timestamp'] >= left_boundary_ts]
if to_timestamp is None and right_boundary_ts:
agg_timeserie = agg_timeserie[
agg_timeserie['timestamp'] <= right_boundary_ts]
points = (agg_timeserie.sort_values(by=['granularity', 'timestamp'],
ascending=[0, 1]).itertuples())
return [(timestamp, granularity, value)
for __, timestamp, granularity, value in points]
class TimeSerieArchive(SerializableMixin):
def __init__(self, agg_timeseries):
"""A raw data buffer and a collection of downsampled timeseries.
Used to represent the set of AggregatedTimeSeries for the range of
granularities supported for a metric (for a particular aggregation
function).
"""
self.agg_timeseries = sorted(agg_timeseries,
key=operator.attrgetter("sampling"))
@classmethod
def from_definitions(cls, definitions, aggregation_method='mean'):
"""Create a new collection of archived time series.
:param definition: A list of tuple (sampling, max_size)
:param aggregation_method: Aggregation function to use.
"""
# Limit the main timeserie to a timespan mapping
return cls(
[AggregatedTimeSerie(
max_size=size,
sampling=sampling,
aggregation_method=aggregation_method)
for sampling, size in definitions]
)
def fetch(self, from_timestamp=None, to_timestamp=None,
timeserie_filter=None):
"""Fetch aggregated time value.
Returns a sorted list of tuples (timestamp, granularity, value).
"""
result = []
end_timestamp = to_timestamp
for ts in reversed(self.agg_timeseries):
if timeserie_filter and not timeserie_filter(ts):
continue
points = ts[from_timestamp:to_timestamp]
try:
# Do not include stop timestamp
del points[end_timestamp]
except KeyError:
pass
result.extend([(timestamp, ts.sampling, value)
for timestamp, value
in six.iteritems(points)])
return result
def update(self, timeserie):
for agg in self.agg_timeseries:
agg.update(timeserie)
def to_dict(self):
return {
"archives": [ts.to_dict() for ts in self.agg_timeseries],
}
def __eq__(self, other):
return (isinstance(other, TimeSerieArchive)
and self.agg_timeseries == other.agg_timeseries)
@classmethod
def from_dict(cls, d):
return cls([AggregatedTimeSerie.from_dict(a) for a in d['archives']])
|