GitLab is used only for code review, issue tracking and project management. Canonical locations for source code are still https://gitweb.torproject.org/ https://git.torproject.org/ and git-rw.torproject.org.

v3bwfile.py 61.9 KB
Newer Older
juga  's avatar
juga committed
1 2 3 4
# -*- coding: utf-8 -*-
"""Classes and functions that create the bandwidth measurements document
(v3bw) used by bandwidth authorities."""

juga  's avatar
juga committed
5
import copy
juga  's avatar
juga committed
6
import logging
7
import math
juga  's avatar
juga committed
8
import os
9
from itertools import combinations
juga  's avatar
juga committed
10
from statistics import median, mean
11
from stem.descriptor import parse_file
juga  's avatar
juga committed
12

13
from sbws import __version__
14
from sbws.globals import (SPEC_VERSION, BW_LINE_SIZE, SBWS_SCALE_CONSTANT,
15 16
                          TORFLOW_SCALING, SBWS_SCALING, TORFLOW_BW_MARGIN,
                          TORFLOW_OBS_LAST, TORFLOW_OBS_MEAN,
17
                          PROP276_ROUND_DIG, MIN_REPORT, MAX_BW_DIFF_PERC)
juga  's avatar
juga committed
18
from sbws.lib.resultdump import ResultSuccess, _ResultType
19
from sbws.util.filelock import DirectoryLock
juga  's avatar
juga committed
20
from sbws.util.timestamp import (now_isodt_str, unixts_to_isodt_str,
21 22
                                 now_unixts, isostr_to_dt_obj,
                                 dt_obj_to_isodt_str)
23
from sbws.util.state import State
juga  's avatar
juga committed
24 25 26

log = logging.getLogger(__name__)

27
LINE_SEP = '\n'
28 29
KEYVALUE_SEP_V1 = '='
KEYVALUE_SEP_V2 = ' '
30 31 32 33 34 35

# NOTE: in a future refactor make make all the KeyValues be a dictionary
# with their type, so that it's more similar to stem parser.

# Header KeyValues
# =================
36 37
# KeyValues that need to be in a specific order in the Bandwidth File.
HEADER_KEYS_V1_1_ORDERED = ['version']
38 39 40 41 42 43 44 45 46
# KeyValues that are not initialized from the state file nor the measurements.
# They can also be pass as an argument to `Header` to overwrite default values,
# what is done in unit tests.
# `latest bandwidth` is special cause it gets its value from timestamp, which
# is not a KeyValue, but it's always pass as an agument.
# It could be separaed in other list, but so far there is no need, cause:
# 1. when it's pass to the Header to initialize it, it's just ignored.
# 2. when the file is created, it's took into account.
HEADER_KEYS_V1_1_SELF_INITIALIZED = [
47 48 49
    "software",
    "software_version",
    "file_created",
50 51 52 53
    "latest_bandwidth",
]
# KeyValues that are initialized from arguments.
HEADER_KEYS_V1_1_TO_INIT = [
54 55
    "earliest_bandwidth",
    "generator_started",
56 57
]

58 59 60 61
# number_eligible_relays is the number that ends in the bandwidth file
# ie, have not been excluded by one of the filters in 4. below
# They should be call recent_measurement_included_count to be congruent
# with the other KeyValues.
62 63 64 65 66 67 68
HEADER_KEYS_V1_2 = [
    "number_eligible_relays",
    "minimum_number_eligible_relays",
    "number_consensus_relays",
    "percent_eligible_relays",
    "minimum_percent_eligible_relays",
]
69

70 71 72 73 74 75
# KeyValues added in the Bandwidth File v1.3.0
HEADER_KEYS_V1_3 = [
    "scanner_country",
    "destinations_countries",
]

76 77 78
# KeyValues that count the number of relays that are in the bandwidth file,
# but ignored by Tor when voting, because they do not have a
# measured bandwidth.
79
HEADER_RECENT_MEASUREMENTS_EXCLUDED_KEYS = [
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
    # Number of relays that were measured but all the measurements failed
    # because of network failures or it was
    # not found a suitable helper relay
    'recent_measurements_excluded_error_count',
    # Number of relays that have successful measurements but the measurements
    # were not away from each other in X time (by default 1 day).
    'recent_measurements_excluded_near_count',
    # Number of relays that have successful measurements and they are away from
    # each other but they are not X time recent.
    # By default this is 5 days, which is the same time the older
    # the measurements can be by default.
    'recent_measurements_excluded_old_count',
    # Number of relays that have successful measurements and they are away from
    # each other and recent
    # but the number of measurements are less than X (by default 2).
    'recent_measurements_excluded_few_count',
]
97
# Added in #29591
98 99 100 101
# NOTE: recent_consensus_count, recent_priority_list_count,
# recent_measurement_attempt_count and recent_priority_relay_count
# are not reset when the scanner is stop.
# They will accumulate the values since the scanner was ever started.
102
HEADER_KEYS_V1_4 = [
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
    # 1.1 header: the number of different consensuses, that sbws has seen,
    # since the last 5 days
    'recent_consensus_count',
    # 2.4 Number of times a priority list has been created
    'recent_priority_list_count',
    # 2.5 Number of relays that there were in a priority list
    # [50, number of relays in the network * 0.05]
    'recent_priority_relay_count',
    # 3.6 header: the number of times that sbws has tried to measure any relay,
    # since the last 5 days
    # This would be the number of times a relays were in a priority list
    'recent_measurement_attempt_count',
    # 3.7 header: the number of times that sbws has tried to measure any relay,
    # since the last 5 days, but it didn't work
    # This should be the number of attempts - number of ResultSuccess -
    # something else we don't know yet
    # So far is the number of ResultError
    'recent_measurement_failure_count',
121 122
    # The time it took to report about half of the network.
    'time_to_report_half_network',
123
] + HEADER_RECENT_MEASUREMENTS_EXCLUDED_KEYS
124

125 126 127 128 129 130 131 132
# KeyValues added in the Bandwidth File v1.5.0
# XXX: Change SPEC_VERSION when all the v1.5.0 keys are added, before a new
# sbws release.
# Tor version will be obtained from the state file, so it won't be pass as an
# argument, but will be self-initialized.
HEADER_KEYS_V1_5_TO_INIT = ['tor_version']
HEADER_KEYS_V1_5 = HEADER_KEYS_V1_5_TO_INIT

133 134 135 136 137
# KeyValues that are initialized from arguments, not self-initialized.
HEADER_INIT_KEYS = (
    HEADER_KEYS_V1_1_TO_INIT
    + HEADER_KEYS_V1_3
    + HEADER_KEYS_V1_2
138
    + HEADER_KEYS_V1_4
139
    + HEADER_KEYS_V1_5_TO_INIT
140
)
141

142
HEADER_INT_KEYS = HEADER_KEYS_V1_2 + HEADER_KEYS_V1_4
143
# List of all unordered KeyValues currently being used to generate the file
144 145 146 147 148 149
HEADER_UNORDERED_KEYS = (
    HEADER_KEYS_V1_1_SELF_INITIALIZED
    + HEADER_KEYS_V1_1_TO_INIT
    + HEADER_KEYS_V1_3
    + HEADER_KEYS_V1_2
    + HEADER_KEYS_V1_4
150
    + HEADER_KEYS_V1_5
151
)
152
# List of all the KeyValues currently being used to generate the file
153
HEADER_ALL_KEYS = HEADER_KEYS_V1_1_ORDERED + HEADER_UNORDERED_KEYS
154

155
TERMINATOR = '====='
156 157 158

# Bandwidth Lines KeyValues
# =========================
159
# Num header lines in v1.X.X using all the KeyValues
160
NUM_LINES_HEADER_V1 = len(HEADER_ALL_KEYS) + 2
161 162
LINE_TERMINATOR = TERMINATOR + LINE_SEP

juga  's avatar
juga committed
163
# KeyValue separator in Bandwidth Lines
164
BWLINE_KEYVALUES_SEP_V1 = ' '
165
# not inclding in the files the extra bws for now
166
BWLINE_KEYS_V0 = ['node_id', 'bw']
167 168 169 170 171 172 173 174 175 176 177 178 179
BWLINE_KEYS_V1_1 = [
    "master_key_ed25519",
    "nick",
    "rtt",
    "time",
    "success",
    "error_stream",
    "error_circ",
    "error_misc",
    # Added in #292951
    "error_second_relay",
    "error_destination",
]
180 181 182 183 184 185 186 187 188 189
BWLINE_KEYS_V1_2 = [
    "bw_median",
    "bw_mean",
    "desc_bw_avg",
    "desc_bw_bur",
    "desc_bw_obs_last",
    "desc_bw_obs_mean",
    "consensus_bandwidth",
    "consensus_bandwidth_is_unmeasured",
]
190

191 192
# There were no bandwidth lines key added in the specification version 1.3

193
# Added in #292951
194
BWLINE_KEYS_V1_4 = [
195 196 197
    # 1.2 relay: the number of different consensuses, that sbws has seen,
    # since the last 5 days, that have this relay
    'relay_in_recent_consensus_count',
198 199 200
    # 2.6 relay: the number of times a relay was "prioritized" to be measured
    # in the recent days (by default 5).
    'relay_recent_priority_list_count',
201 202 203 204 205 206 207 208 209 210 211 212
    # 3.8 relay:  the number of times that sbws has tried to measure
    # this relay, since the last 5 days
    # This would be the number of times a relay was in a priority list (2.6)
    # since once it gets measured, it either returns ResultError,
    # ResultSuccess or something else happened that we don't know yet
    'relay_recent_measurement_attempt_count',
    # 3.9 relay:  the number of times that sbws has tried to measure
    # this relay, since the last 5 days, but it didn't work
    # This should be the number of attempts - number of ResultSuccess -
    # something else we don't know yet
    # So far is the number of ResultError
    'relay_recent_measurement_failure_count',
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
    # Number of error results created in the last 5 days that are excluded.
    # This is the sum of all the errors.
    'relay_recent_measurements_excluded_error_count',
    # The number of successful results, created in the last 5 days,
    # that were excluded by a rule, for this relay.
    # 'relay_recent_measurements_excluded_error_count' would be the
    # sum of the following 3 + the number of error results.

    # The number of successful measurements that are not X time away
    # from each other (by default 1 day).
    'relay_recent_measurements_excluded_near_count',
    # The number of successful measurements that are away from each other
    # but not X time recent (by default 5 days).
    'relay_recent_measurements_excluded_old_count',
    # The number of measurements excluded because they are not at least X
    # (by default 2).
    'relay_recent_measurements_excluded_few_count',
230 231 232 233 234 235 236 237 238 239 240 241 242
    # `vote=0` is used for the relays that were excluded to
    # be reported in the bandwidth file and now they are
    # reported.
    # It tells Tor to do not vote on the relay.
    # `unmeasured=1` is used for the same relays and it is
    # added in case Tor would vote on them in future versions.
    # Maybe these keys should not be included for the relays
    # in which vote=1 and unmeasured=0.
    'vote', 'unmeasured',
    # When there not enough eligible relays (not excluded)
    # under_min_report is 1, `vote` is 0.
    # Added in #29853.
    'under_min_report',
243
]
244
BWLINE_KEYS_V1 = BWLINE_KEYS_V0 + BWLINE_KEYS_V1_1 + BWLINE_KEYS_V1_2 \
245
               + BWLINE_KEYS_V1_4
246 247
# NOTE: tech-debt: assign boolean type to vote and unmeasured,
# when the attributes are defined with a type, as stem does.
248 249 250 251 252 253 254 255 256 257 258 259
BWLINE_INT_KEYS = (
    [
        "bw",
        "rtt",
        "success",
        "error_stream",
        "error_circ",
        "error_misc",
    ]
    + BWLINE_KEYS_V1_2
    + BWLINE_KEYS_V1_4
)
juga  's avatar
juga committed
260 261


262
def round_sig_dig(n, digits=PROP276_ROUND_DIG):
263 264 265 266 267 268 269
    """Round n to 'digits' significant digits in front of the decimal point.
       Results less than or equal to 1 are rounded to 1.
       Returns an integer.

       digits must be greater than 0.
       n must be less than or equal to 2**73, to avoid floating point errors.
       """
270
    digits = int(digits)
271 272 273 274 275 276 277 278 279
    assert digits >= 1
    if n <= 1:
        return 1
    digits_in_n = int(math.log10(n)) + 1
    round_digits = max(digits_in_n - digits, 0)
    rounded_n = round(n, -round_digits)
    return int(rounded_n)


280
def kb_round_x_sig_dig(bw_bs, digits=PROP276_ROUND_DIG):
281 282 283 284 285 286 287 288 289 290 291
    """Convert bw_bs from bytes to kilobytes, and round the result to
       'digits' significant digits.
       Results less than or equal to 1 are rounded up to 1.
       Returns an integer.

       digits must be greater than 0.
       n must be less than or equal to 2**82, to avoid floating point errors.
       """
    # avoid double-rounding by using floating-point
    bw_kb = bw_bs / 1000.0
    return round_sig_dig(bw_kb, digits=digits)
292 293


juga  's avatar
juga committed
294 295 296 297 298 299 300 301 302
def num_results_of_type(results, type_str):
    return len([r for r in results if r.type == type_str])


# Better way to use enums?
def result_type_to_key(type_str):
    return type_str.replace('-', '_')


303
class V3BWHeader(object):
juga  's avatar
juga committed
304 305
    """
    Create a bandwidth measurements (V3bw) header
306
    following bandwidth measurements document spec version 1.X.X.
juga  's avatar
juga committed
307

308
    :param str timestamp: timestamp in Unix Epoch seconds of the most recent
309
        generator result.
juga  's avatar
juga committed
310 311 312
    :param str version: the spec version
    :param str software: the name of the software that generates this
    :param str software_version: the version of the software
313
    :param dict kwargs: extra headers. Currently supported:
juga  's avatar
juga committed
314

315 316 317 318
        - earliest_bandwidth: str, ISO 8601 timestamp in UTC time zone
          when the first bandwidth was obtained
        - generator_started: str, ISO 8601 timestamp in UTC time zone
          when the generator started
juga  's avatar
juga committed
319
    """
320
    def __init__(self, timestamp, **kwargs):
juga  's avatar
juga committed
321 322 323 324
        assert isinstance(timestamp, str)
        for v in kwargs.values():
            assert isinstance(v, str)
        self.timestamp = timestamp
325 326 327 328
        # KeyValues with default value when not given by kwargs
        self.version = kwargs.get('version', SPEC_VERSION)
        self.software = kwargs.get('software', 'sbws')
        self.software_version = kwargs.get('software_version', __version__)
329
        self.file_created = kwargs.get('file_created', now_isodt_str())
juga  's avatar
juga committed
330
        # latest_bandwidth should not be in kwargs, since it MUST be the
331
        # same as timestamp
juga  's avatar
juga committed
332
        self.latest_bandwidth = unixts_to_isodt_str(timestamp)
333
        [setattr(self, k, v) for k, v in kwargs.items()
334
         if k in HEADER_INIT_KEYS]
335

336
    def __str__(self):
337
        if self.version.startswith('1.'):
338 339
            return self.strv1
        return self.strv2
340

juga  's avatar
juga committed
341
    @classmethod
342 343
    def from_results(cls, results, scanner_country=None,
                     destinations_countries=None, state_fpath=''):
juga  's avatar
juga committed
344 345 346
        kwargs = dict()
        latest_bandwidth = cls.latest_bandwidth_from_results(results)
        earliest_bandwidth = cls.earliest_bandwidth_from_results(results)
347
        # NOTE: Blocking, reads file
348
        generator_started = cls.generator_started_from_file(state_fpath)
349
        recent_consensus_count = cls.consensus_count_from_file(state_fpath)
juga  's avatar
juga committed
350
        timestamp = str(latest_bandwidth)
351 352 353 354 355 356 357 358 359

        # XXX: tech-debt: obtain the other values from the state file using
        # this state variable.
        # Store the state as an attribute of the object?
        state = State(state_fpath)
        tor_version = state.get('tor_version', None)
        if tor_version:
            kwargs['tor_version'] = tor_version

juga  's avatar
juga committed
360 361 362 363
        kwargs['latest_bandwidth'] = unixts_to_isodt_str(latest_bandwidth)
        kwargs['earliest_bandwidth'] = unixts_to_isodt_str(earliest_bandwidth)
        if generator_started is not None:
            kwargs['generator_started'] = generator_started
364 365 366
        # To be compatible with older bandwidth files, do not require it.
        if scanner_country is not None:
            kwargs['scanner_country'] = scanner_country
367 368
        if destinations_countries is not None:
            kwargs['destinations_countries'] = destinations_countries
369
        if recent_consensus_count is not None:
370
            kwargs['recent_consensus_count'] = recent_consensus_count
371 372 373 374 375 376 377 378 379 380 381

        recent_measurement_attempt_count = \
            cls.recent_measurement_attempt_count_from_file(state_fpath)
        if recent_measurement_attempt_count is not None:
            kwargs['recent_measurement_attempt_count'] = \
                str(recent_measurement_attempt_count)

        # If it is a failure that is not a ResultError, then
        # failures = attempts - all mesaurements
        # Works only in the case that old measurements files already had
        # measurements count
382 383
        # If this is None or 0, the failures can't be calculated
        if recent_measurement_attempt_count:
384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
            all_measurements = 0
            for result_list in results.values():
                all_measurements += len(result_list)
            measurement_failures = (recent_measurement_attempt_count
                                    - all_measurements)
            kwargs['recent_measurement_failure_count'] = \
                str(measurement_failures)

        priority_lists = cls.recent_priority_list_count_from_file(state_fpath)
        if priority_lists is not None:
            kwargs['recent_priority_list_count'] = str(priority_lists)

        priority_relays = \
            cls.recent_priority_relay_count_from_file(state_fpath)
        if priority_relays is not None:
            kwargs['recent_priority_relay_count'] = str(priority_relays)

juga  's avatar
juga committed
401 402 403
        h = cls(timestamp, **kwargs)
        return h

404
    @classmethod
405
    def from_lines_v1(cls, lines):
406 407
        """
        :param list lines: list of lines to parse
408
        :returns: tuple of V3BWHeader object and non-header lines
409 410 411 412
        """
        assert isinstance(lines, list)
        try:
            index_terminator = lines.index(TERMINATOR)
Matt Traudt's avatar
Matt Traudt committed
413
        except ValueError:
414 415 416
            # is not a bw file or is v100
            log.warn('Terminator is not in lines')
            return None
417
        ts = lines[0]
418
        kwargs = dict([l.split(KEYVALUE_SEP_V1)
419
                       for l in lines[:index_terminator]
420
                       if l.split(KEYVALUE_SEP_V1)[0] in HEADER_ALL_KEYS])
421
        h = cls(ts, **kwargs)
juga  's avatar
juga committed
422
        # last line is new line
423
        return h, lines[index_terminator + 1:-1]
424 425

    @classmethod
426
    def from_text_v1(self, text):
427
        """
428
        :param str text: text to parse
429
        :returns: tuple of V3BWHeader object and non-header lines
430 431
        """
        assert isinstance(text, str)
432
        return self.from_lines_v1(text.split(LINE_SEP))
juga  's avatar
juga committed
433

juga  's avatar
juga committed
434 435 436 437 438 439 440 441 442 443 444
    @classmethod
    def from_lines_v100(cls, lines):
        """
        :param list lines: list of lines to parse
        :returns: tuple of V3BWHeader object and non-header lines
        """
        assert isinstance(lines, list)
        h = cls(lines[0])
        # last line is new line
        return h, lines[1:-1]

juga  's avatar
juga committed
445
    @staticmethod
446
    def generator_started_from_file(state_fpath):
447 448 449 450
        '''
        ISO formatted timestamp for the time when the scanner process most
        recently started.
        '''
451
        state = State(state_fpath)
452
        if 'scanner_started' in state:
453 454
            # From v1.1.0-dev `state` is capable of converting strs to datetime
            return dt_obj_to_isodt_str(state['scanner_started'])
455 456
        else:
            return None
juga  's avatar
juga committed
457

458 459 460
    @staticmethod
    def consensus_count_from_file(state_fpath):
        state = State(state_fpath)
461 462 463 464
        count = state.count("recent_consensus")
        if count:
            return str(count)
        return None
465

466 467 468 469 470 471 472 473
    # NOTE: in future refactor store state in the class
    @staticmethod
    def recent_measurement_attempt_count_from_file(state_fpath):
        """
        Returns the number of times any relay was queued to be measured
        in the recent (by default 5) days from the state file.
        """
        state = State(state_fpath)
474
        return state.count('recent_measurement_attempt')
475 476 477 478 479 480 481 482 483 484

    @staticmethod
    def recent_priority_list_count_from_file(state_fpath):
        """
        Returns the number of times
        :meth:`~sbws.lib.relayprioritizer.RelayPrioritizer.best_priority`
        was run
        in the recent (by default 5) days from the state file.
        """
        state = State(state_fpath)
485
        return state.count('recent_priority_list')
486 487 488 489 490 491 492 493

    @staticmethod
    def recent_priority_relay_count_from_file(state_fpath):
        """
        Returns the number of times any relay was "prioritized" to be measured
        in the recent (by default 5) days from the state file.
        """
        state = State(state_fpath)
494
        return state.count('recent_priority_relay')
495

juga  's avatar
juga committed
496
    @staticmethod
juga  's avatar
juga committed
497
    def latest_bandwidth_from_results(results):
498
        return round(max([r.time for fp in results for r in results[fp]]))
juga  's avatar
juga committed
499 500 501

    @staticmethod
    def earliest_bandwidth_from_results(results):
502
        return round(min([r.time for fp in results for r in results[fp]]))
juga  's avatar
juga committed
503

juga  's avatar
juga committed
504 505 506 507 508
    @property
    def keyvalue_unordered_tuple_ls(self):
        """Return list of KeyValue tuples that do not have specific order."""
        # sort the list to generate determinist headers
        keyvalue_tuple_ls = sorted([(k, v) for k, v in self.__dict__.items()
509
                                    if k in HEADER_UNORDERED_KEYS])
juga  's avatar
juga committed
510 511 512 513 514 515 516 517
        return keyvalue_tuple_ls

    @property
    def keyvalue_tuple_ls(self):
        """Return list of all KeyValue tuples"""
        return [('version', self.version)] + self.keyvalue_unordered_tuple_ls

    @property
518 519 520
    def keyvalue_v1str_ls(self):
        """Return KeyValue list of strings following spec v1.X.X."""
        keyvalues = [self.timestamp] + [KEYVALUE_SEP_V1.join([k, v])
juga  's avatar
juga committed
521 522 523 524
                                        for k, v in self.keyvalue_tuple_ls]
        return keyvalues

    @property
525 526 527
    def strv1(self):
        """Return header string following spec v1.X.X."""
        header_str = LINE_SEP.join(self.keyvalue_v1str_ls) + LINE_SEP + \
juga  's avatar
juga committed
528 529 530 531
            LINE_TERMINATOR
        return header_str

    @property
532 533 534
    def keyvalue_v2_ls(self):
        """Return KeyValue list of strings following spec v2.X.X."""
        keyvalue = [self.timestamp] + [KEYVALUE_SEP_V2.join([k, v])
juga  's avatar
juga committed
535 536 537 538
                                       for k, v in self.keyvalue_tuple_ls]
        return keyvalue

    @property
539 540 541
    def strv2(self):
        """Return header string following spec v2.X.X."""
        header_str = LINE_SEP.join(self.keyvalue_v2_ls) + LINE_SEP + \
juga  's avatar
juga committed
542 543 544 545 546 547
            LINE_TERMINATOR
        return header_str

    @property
    def num_lines(self):
        return len(self.__str__().split(LINE_SEP))
juga  's avatar
juga committed
548

juga  's avatar
juga committed
549 550 551
    def add_stats(self, **kwargs):
        # Using kwargs because attributes might chage.
        [setattr(self, k, str(v)) for k, v in kwargs.items()
552
         if k in HEADER_KEYS_V1_2]
juga  's avatar
juga committed
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
    def add_time_report_half_network(self):
        """Add to the header the time it took to measure half of the network.

        It is not the time the scanner actually takes on measuring all the
        network, but the ``number_eligible_relays`` that are reported in the
        bandwidth file and directory authorities will vote on.

        This is calculated for half of the network, so that failed or not
        reported relays do not affect too much.

        For instance, if there are 6500 relays in the network, half of the
        network would be 3250. And if there were 4000 eligible relays
        measured in an interval of 3 days, the time to measure half of the
        network would be 3 days * 3250 / 4000.

        Since the elapsed time is calculated from the earliest and the
        latest measurement and a relay might have more than 2 measurements,
        this would give an estimate on how long it would take to measure
        the network including all the valid measurements.

        Log also an estimated on how long it would take with the current
        number of relays included in the bandwidth file.
        """
        # NOTE: in future refactor do not convert attributes to str until
        # writing to the file, so that they do not need to be converted back
        # to do some calculations.
        elapsed_time = (
            (isostr_to_dt_obj(self.latest_bandwidth)
             - isostr_to_dt_obj(self.earliest_bandwidth))
            .total_seconds())

        # This attributes were added later and some tests that
        # do not initialize them would fail.
        eligible_relays = int(getattr(self, 'number_eligible_relays', 0))
        consensus_relays = int(getattr(self, 'number_consensus_relays', 0))
        if not(eligible_relays and consensus_relays):
            return

        half_network = consensus_relays / 2
        # Calculate the time it would take to measure half of the network
        if eligible_relays >= half_network:
            time_half_network = round(
                elapsed_time * half_network / eligible_relays
            )
            self.time_to_report_half_network = str(time_half_network)

        # In any case log an estimated on the time to measure all the network.
        estimated_time = round(
            elapsed_time * consensus_relays / eligible_relays
        )
        log.info("Estimated time to measure the network: %s hours.",
                 round(estimated_time / 60 / 60))

607 608 609 610 611 612 613 614 615
    def add_relays_excluded_counters(self, exclusion_dict):
        """
        Add the monitoring KeyValues to the header about the number of
        relays not included because they were not ``eligible``.
        """
        log.debug("Adding relays excluded counters.")
        for k, v in exclusion_dict.items():
            setattr(self, k, str(v))

juga  's avatar
juga committed
616

juga  's avatar
juga committed
617
class V3BWLine(object):
juga  's avatar
juga committed
618
    """
619
    Create a Bandwidth List line following the spec version 1.X.X.
juga  's avatar
juga committed
620

621 622 623 624
    :param str node_id: the relay fingerprint
    :param int bw: the bandwidth value that directory authorities will include
        in their votes.
    :param dict kwargs: extra headers.
juga  's avatar
juga committed
625

626 627
    .. note:: tech-debt: move node_id and bw to kwargs and just ensure that
       the required values are in **kwargs
juga  's avatar
juga committed
628
    """
juga  's avatar
juga committed
629 630
    def __init__(self, node_id, bw, **kwargs):
        assert isinstance(node_id, str)
631
        assert node_id.startswith('$')
juga  's avatar
juga committed
632 633 634
        self.node_id = node_id
        self.bw = bw
        [setattr(self, k, v) for k, v in kwargs.items()
635
         if k in BWLINE_KEYS_V1]
juga  's avatar
juga committed
636

juga  's avatar
juga committed
637
    def __str__(self):
638
        return self.bw_strv1
juga  's avatar
juga committed
639

juga  's avatar
juga committed
640
    @classmethod
juga  's avatar
juga committed
641 642
    def from_results(cls, results, secs_recent=None, secs_away=None,
                     min_num=0):
643 644 645
        """Convert sbws results to relays' Bandwidth Lines

        ``bs`` stands for Bytes/seconds
646
        ``bw_mean`` means the bw is obtained from the mean of the all the
647 648 649 650 651
        downloads' bandwidth.
        Downloads' bandwidth are calculated as the amount of data received
        divided by the the time it took to received.
        bw = data (Bytes) / time (seconds)
        """
652
        # log.debug("Len success_results %s", len(success_results))
juga  's avatar
juga committed
653 654 655 656 657 658 659
        node_id = '$' + results[0].fingerprint
        kwargs = dict()
        kwargs['nick'] = results[0].nickname
        if getattr(results[0], 'master_key_ed25519'):
            kwargs['master_key_ed25519'] = results[0].master_key_ed25519
        kwargs['time'] = cls.last_time_from_results(results)
        kwargs.update(cls.result_types_from_results(results))
660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682

        # If it has not the attribute, return list to be able to call len
        # If it has the attribute, but it is None, return also list
        kwargs['relay_in_recent_consensus_count'] = str(
            max([
                len(getattr(r, 'relay_in_recent_consensus', []) or [])
                for r in results
            ])
        )

        kwargs['relay_recent_priority_list_count'] = str(
            max([
                len(getattr(r, 'relay_recent_priority_list', []) or [])
                for r in results
            ])
        )

        kwargs['relay_recent_measurement_attempt_count'] = str(
            max([
                len(getattr(r, 'relay_recent_measurement_attempt', []) or [])
                for r in results
            ])
        )
683

684
        success_results = [r for r in results if isinstance(r, ResultSuccess)]
685 686 687 688

        # NOTE: The following 4 conditions exclude relays from the bandwidth
        # file when the measurements does not satisfy some rules, what makes
        # the relay non-`eligible`.
689 690
        # In BWLINE_KEYS_V1_4 it is explained what they mean.
        # In HEADER_RECENT_MEASUREMENTS_EXCLUDED_KEYS it is also
691 692 693 694 695 696
        # explained the what it means the strings returned.
        # They rules were introduced in #28061 and #27338
        # In #28565 we introduce the KeyValues to know why they're excluded.
        # In #28563 we report these relays, but make Tor ignore them.
        # This might confirm #28042.

697 698 699 700 701 702 703 704 705 706 707 708 709 710
        # If the relay is non-`eligible`:
        # Create a bandwidth line with the relay, but set ``vote=0`` so that
        # Tor versions with patch #29806 does not vote on the relay.
        # Set ``bw=1`` so that Tor versions without the patch,
        # will give the relay low bandwidth.
        # Include ``unmeasured=1`` in case Tor would vote on unmeasured relays
        # in future versions.
        # And return because there are not bandwidth values.
        # NOTE: the bandwidth values could still be obtained if:
        # 1. ``ResultError`` will store them
        # 2. assign ``results_recent = results`` when there is a ``exclusion
        # reason.
        # This could be done in a better way as part of a refactor #28684.

juga  's avatar
juga committed
711 712
        kwargs['vote'] = 0
        kwargs['unmeasured'] = 1
713 714

        exclusion_reason = None
715

716 717 718 719 720
        number_excluded_error = len(results) - len(success_results)
        if number_excluded_error > 0:
            # then the number of error results is the number of results
            kwargs['relay_recent_measurements_excluded_error_count'] = \
                number_excluded_error
721
        if not success_results:
722
            exclusion_reason = 'recent_measurements_excluded_error_count'
723
            return (cls(node_id, 1, **kwargs), exclusion_reason)
724

725 726
        results_away = \
            cls.results_away_each_other(success_results, secs_away)
727 728 729
        number_excluded_near = len(success_results) - len(results_away)
        if number_excluded_near > 0:
            kwargs['relay_recent_measurements_excluded_near_count'] = \
730
                number_excluded_near
731
        if not results_away:
732
            exclusion_reason = \
733
                'recent_measurements_excluded_near_count'
734
            return (cls(node_id, 1, **kwargs), exclusion_reason)
735 736
        # log.debug("Results away from each other: %s",
        #           [unixts_to_isodt_str(r.time) for r in results_away])
737

738
        results_recent = cls.results_recent_than(results_away, secs_recent)
739 740 741 742
        number_excluded_old = len(results_away) - len(results_recent)
        if number_excluded_old > 0:
            kwargs['relay_recent_measurements_excluded_old_count'] = \
                number_excluded_old
743
        if not results_recent:
744
            exclusion_reason = \
745
                'recent_measurements_excluded_old_count'
746
            return (cls(node_id, 1, **kwargs), exclusion_reason)
747

748
        if not len(results_recent) >= min_num:
749 750
            kwargs['relay_recent_measurements_excluded_few_count'] = \
                len(results_recent)
751
            # log.debug('The number of results is less than %s', min_num)
752
            exclusion_reason = \
753
                'recent_measurements_excluded_few_count'
754 755 756 757 758 759
            return (cls(node_id, 1, **kwargs), exclusion_reason)

        # For any line not excluded, do not include vote and unmeasured
        # KeyValues
        del kwargs['vote']
        del kwargs['unmeasured']
760

761 762 763 764 765 766 767 768 769 770 771 772 773 774 775
        rtt = cls.rtt_from_results(results_recent)
        if rtt:
            kwargs['rtt'] = rtt
        bw = cls.bw_median_from_results(results_recent)
        kwargs['bw_mean'] = cls.bw_mean_from_results(results_recent)
        kwargs['bw_median'] = cls.bw_median_from_results(
            results_recent)
        kwargs['desc_bw_avg'] = \
            cls.desc_bw_avg_from_results(results_recent)
        kwargs['desc_bw_bur'] = \
            cls.desc_bw_bur_from_results(results_recent)
        kwargs['consensus_bandwidth'] = \
            cls.consensus_bandwidth_from_results(results_recent)
        kwargs['consensus_bandwidth_is_unmeasured'] = \
            cls.consensus_bandwidth_is_unmeasured_from_results(
776
                results_recent)
777 778 779 780 781
        kwargs['desc_bw_obs_last'] = \
            cls.desc_bw_obs_last_from_results(results_recent)
        kwargs['desc_bw_obs_mean'] = \
            cls.desc_bw_obs_mean_from_results(results_recent)
        bwl = cls(node_id, bw, **kwargs)
782
        return bwl, None
juga  's avatar
juga committed
783 784 785 786 787 788

    @classmethod
    def from_data(cls, data, fingerprint):
        assert fingerprint in data
        return cls.from_results(data[fingerprint])

juga  's avatar
juga committed
789
    @classmethod
790
    def from_bw_line_v1(cls, line):
juga  's avatar
juga committed
791
        assert isinstance(line, str)
792
        kwargs = dict([kv.split(KEYVALUE_SEP_V1)
793
                       for kv in line.split(BWLINE_KEYVALUES_SEP_V1)
794
                       if kv.split(KEYVALUE_SEP_V1)[0] in BWLINE_KEYS_V1])
juga  's avatar
juga committed
795
        for k, v in kwargs.items():
796
            if k in BWLINE_INT_KEYS:
juga  's avatar
juga committed
797
                kwargs[k] = int(v)
798 799 800 801 802
        node_id = kwargs['node_id']
        bw = kwargs['bw']
        del kwargs['node_id']
        del kwargs['bw']
        bw_line = cls(node_id, bw, **kwargs)
juga  's avatar
juga committed
803 804
        return bw_line

juga  's avatar
juga committed
805 806
    @staticmethod
    def results_away_each_other(results, secs_away=None):
juga  's avatar
juga committed
807 808
        # log.debug("Checking whether results are away from each other in %s "
        #           "secs.", secs_away)
juga  's avatar
juga committed
809 810
        if secs_away is None or len(results) < 2:
            return results
811 812 813 814 815
        for a, b in combinations(results, 2):
            if abs(a.time - b.time) > secs_away:
                return results
        # log.debug("Results are NOT away from each other in at least %ss: %s",
        #           secs_away, [unixts_to_isodt_str(r.time) for r in results])
816
        return []
juga  's avatar
juga committed
817 818 819 820 821

    @staticmethod
    def results_recent_than(results, secs_recent=None):
        if secs_recent is None:
            return results
822
        results_recent = list(filter(
juga  's avatar
juga committed
823
                            lambda x: (now_unixts() - x.time) < secs_recent,
824 825 826 827 828 829
                            results))
        # if not results_recent:
        #     log.debug("Results are NOT more recent than %ss: %s",
        #               secs_recent,
        #               [unixts_to_isodt_str(r.time) for r in results])
        return results_recent
juga  's avatar
juga committed
830

831
    @staticmethod
832
    def bw_median_from_results(results):
833 834 835 836
        return max(round(median([dl['amount'] / dl['duration']
                                 for r in results for dl in r.downloads])), 1)

    @staticmethod
837
    def bw_mean_from_results(results):
838 839 840
        return max(round(mean([dl['amount'] / dl['duration']
                               for r in results for dl in r.downloads])), 1)

juga  's avatar
juga committed
841 842 843 844 845 846 847 848
    @staticmethod
    def last_time_from_results(results):
        return unixts_to_isodt_str(round(max([r.time for r in results])))

    @staticmethod
    def rtt_from_results(results):
        # convert from miliseconds to seconds
        rtts = [(round(rtt * 1000)) for r in results for rtt in r.rtts]
juga  's avatar
juga committed
849
        rtt = round(median(rtts)) if rtts else None
juga  's avatar
juga committed
850 851 852 853 854 855 856 857 858
        return rtt

    @staticmethod
    def result_types_from_results(results):
        rt_dict = dict([(result_type_to_key(rt.value),
                         num_results_of_type(results, rt.value))
                        for rt in _ResultType])
        return rt_dict

859 860 861 862 863 864 865 866
    @staticmethod
    def desc_bw_avg_from_results(results):
        """Obtain the last descriptor bandwidth average from the results."""
        for r in reversed(results):
            if r.relay_average_bandwidth is not None:
                return r.relay_average_bandwidth
        return None

867 868 869 870 871 872 873 874
    @staticmethod
    def desc_bw_bur_from_results(results):
        """Obtain the last descriptor bandwidth burst from the results."""
        for r in reversed(results):
            if r.relay_burst_bandwidth is not None:
                return r.relay_burst_bandwidth
        return None

875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890
    @staticmethod
    def consensus_bandwidth_from_results(results):
        """Obtain the last consensus bandwidth from the results."""
        for r in reversed(results):
            if r.consensus_bandwidth is not None:
                return r.consensus_bandwidth
        return None

    @staticmethod
    def consensus_bandwidth_is_unmeasured_from_results(results):
        """Obtain the last consensus unmeasured flag from the results."""
        for r in reversed(results):
            if r.consensus_bandwidth_is_unmeasured is not None:
                return r.consensus_bandwidth_is_unmeasured
        return None

891
    @staticmethod
892 893
    def desc_bw_obs_mean_from_results(results):
        desc_bw_obs_ls = []
894 895
        for r in results:
            if r.relay_observed_bandwidth is not None:
896 897 898
                desc_bw_obs_ls.append(r.relay_observed_bandwidth)
        if desc_bw_obs_ls:
            return max(round(mean(desc_bw_obs_ls)), 1)
899 900 901
        return None

    @staticmethod
902
    def desc_bw_obs_last_from_results(results):
903 904 905 906 907 908
        # the last is at the end of the list
        for r in reversed(results):
            if r.relay_observed_bandwidth is not None:
                return r.relay_observed_bandwidth
        return None

juga  's avatar
juga committed
909 910 911 912 913
    @property
    def bw_keyvalue_tuple_ls(self):
        """Return list of KeyValue Bandwidth Line tuples."""
        # sort the list to generate determinist headers
        keyvalue_tuple_ls = sorted([(k, v) for k, v in self.__dict__.items()
914
                                    if k in BWLINE_KEYS_V1])
juga  's avatar
juga committed
915
        return keyvalue_tuple_ls
juga  's avatar
juga committed
916

juga  's avatar
juga committed
917
    @property
918
    def bw_keyvalue_v1str_ls(self):
juga  's avatar
juga committed
919
        """Return list of KeyValue Bandwidth Line strings following
920
        spec v1.X.X.
juga  's avatar
juga committed
921
        """
922
        bw_keyvalue_str = [KEYVALUE_SEP_V1 .join([k, str(v)])
juga  's avatar
juga committed
923 924
                           for k, v in self.bw_keyvalue_tuple_ls]
        return bw_keyvalue_str
juga  's avatar
juga committed
925

juga  's avatar
juga committed
926
    @property
927 928
    def bw_strv1(self):
        """Return Bandwidth Line string following spec v1.X.X."""
929
        bw_line_str = BWLINE_KEYVALUES_SEP_V1.join(
930
                        self.bw_keyvalue_v1str_ls) + LINE_SEP
juga  's avatar
juga committed
931 932 933 934 935 936
        if len(bw_line_str) > BW_LINE_SIZE:
            # if this is the case, probably there are too many KeyValues,
            # or the limit needs to be changed in Tor
            log.warn("The bandwidth line %s is longer than %s",
                     len(bw_line_str), BW_LINE_SIZE)
        return bw_line_str
juga  's avatar
juga committed
937

juga  's avatar
juga committed
938

939
class V3BWFile(object):
juga  's avatar
juga committed
940
    """
941
    Create a Bandwidth List file following spec version 1.X.X
juga  's avatar
juga committed
942 943 944 945

    :param V3BWHeader v3bwheader: header
    :param list v3bwlines: V3BWLines
    """
juga  's avatar
juga committed
946 947 948 949 950
    def __init__(self, v3bwheader, v3bwlines):
        self.header = v3bwheader
        self.bw_lines = v3bwlines

    def __str__(self):
951
        return str(self.header) + ''.join([str(bw_line) or ''
juga  's avatar
juga committed
952 953
                                           for bw_line in self.bw_lines])

juga  's avatar
juga committed
954
    @classmethod
955 956
    def from_results(cls, results, scanner_country=None,
                     destinations_countries=None, state_fpath='',
957
                     scale_constant=SBWS_SCALE_CONSTANT,
958 959
                     scaling_method=TORFLOW_SCALING,
                     torflow_obs=TORFLOW_OBS_LAST,
juga  's avatar
juga committed
960
                     torflow_cap=TORFLOW_BW_MARGIN,
961
                     round_digs=PROP276_ROUND_DIG,
962
                     secs_recent=None, secs_away=None, min_num=0,
963 964
                     consensus_path=None, max_bw_diff_perc=MAX_BW_DIFF_PERC,
                     reverse=False):
965 966 967 968 969 970
        """Create V3BWFile class from sbws Results.

        :param dict results: see below
        :param str state_fpath: path to the state file
        :param int scaling_method:
            Scaling method to obtain the bandwidth
971
            Possible values: {None, SBWS_SCALING, TORFLOW_SCALING} = {0, 1, 2}
972 973 974 975 976 977 978 979 980 981 982
        :param int scale_constant: sbws scaling constant
        :param int torflow_obs: method to choose descriptor observed bandwidth
        :param bool reverse: whether to sort the bw lines descending or not

        Results are in the form::

            {'relay_fp1': [Result1, Result2, ...],
             'relay_fp2': [Result1, Result2, ...]}

        """
        log.info('Processing results to generate a bandwidth list file.')
983 984
        header = V3BWHeader.from_results(results, scanner_country,
                                         destinations_countries, state_fpath)
985
        bw_lines_raw = []
986
        bw_lines_excluded = []
juga  's avatar
juga committed
987 988
        number_consensus_relays = cls.read_number_consensus_relays(
            consensus_path)
989
        state = State(state_fpath)
990 991 992

        # Create a dictionary with the number of relays excluded by any of the
        # of the filtering rules that makes relays non-`eligible`.
993
        # NOTE: In HEADER_RECENT_MEASUREMENTS_EXCLUDED_KEYS it is
994 995 996
        # explained what are the KeyValues.
        # See also the comments in `from_results`.
        exclusion_dict = dict(
997
            [(k, 0) for k in HEADER_RECENT_MEASUREMENTS_EXCLUDED_KEYS]
998
            )
juga  's avatar
juga committed
999
        for fp, values in results.items():
1000
            # log.debug("Relay fp %s", fp)
1001 1002
            line, reason = V3BWLine.from_results(values, secs_recent,
                                                 secs_away, min_num)
1003 1004
            # If there is no reason it means the line will not be excluded.
            if not reason:
juga  's avatar
juga committed
1005
                bw_lines_raw.append(line)
1006
            else:
1007 1008 1009
                # Store the excluded lines to include them in the bandwidth
                # file.
                bw_lines_excluded.append(line)
1010 1011 1012 1013
                exclusion_dict[reason] = exclusion_dict.get(reason, 0) + 1
        # Add the headers with the number of excluded relays by reason
        header.add_relays_excluded_counters(exclusion_dict)

1014
        if not bw_lines_raw:
1015 1016
            # It could be possible to scale the lines that were successful
            # even if excluded, but is not done here.
1017 1018
            log.info("After applying restrictions to the raw results, "
                     "there is not any. Scaling can not be applied.")
1019
            # Update the header and log the progress.
juga  's avatar
juga committed
1020
            cls.update_progress(
1021 1022 1023 1024 1025
                cls, 0, header, number_consensus_relays, state)
            # Set the lines that would be excluded anyway (`vote=0`) with
            # `under_min_report=1`
            cls.set_under_min_report(bw_lines_excluded)
            # Create the bandwidth file with the lines that would be excluded.
1026
            return cls(header, bw_lines_excluded)
1027 1028
        if scaling_method == SBWS_SCALING:
            bw_lines = cls.bw_sbws_scale(bw_lines_raw, scale_constant)