xref: /openbmc/openbmc/poky/scripts/oe-build-perf-report (revision edff49234e31f23dc79f823473c9e286a21596c1)
1#!/usr/bin/env python3
2#
3# Examine build performance test results
4#
5# Copyright (c) 2017, Intel Corporation.
6#
7# SPDX-License-Identifier: GPL-2.0-only
8#
9
10import argparse
11import json
12import logging
13import os
14import re
15import sys
16from collections import namedtuple, OrderedDict
17from operator import attrgetter
18from xml.etree import ElementTree as ET
19
20# Import oe libs
21scripts_path = os.path.dirname(os.path.realpath(__file__))
22sys.path.append(os.path.join(scripts_path, 'lib'))
23import scriptpath
24from build_perf import print_table
25from build_perf.report import (metadata_xml_to_json, results_xml_to_json,
26                               aggregate_data, aggregate_metadata, measurement_stats,
27                               AggregateTestData)
28from build_perf import html
29from buildstats import BuildStats, diff_buildstats, BSVerDiff
30
31scriptpath.add_oe_lib_path()
32
33from oeqa.utils.git import GitRepo, GitError
34import oeqa.utils.gitarchive as gitarchive
35
36
37# Setup logging
38logging.basicConfig(level=logging.INFO, format="%(levelname)s: %(message)s")
39log = logging.getLogger('oe-build-perf-report')
40
41def list_test_revs(repo, tag_name, verbosity, **kwargs):
42    """Get list of all tested revisions"""
43    valid_kwargs = dict([(k, v) for k, v in kwargs.items() if v is not None])
44
45    fields, revs = gitarchive.get_test_runs(log, repo, tag_name, **valid_kwargs)
46    ignore_fields = ['tag_number']
47    if verbosity < 2:
48        extra_fields = ['COMMITS', 'TEST RUNS']
49        ignore_fields.extend(['commit_number', 'commit'])
50    else:
51        extra_fields = ['TEST RUNS']
52
53    print_fields = [i for i, f in enumerate(fields) if f not in ignore_fields]
54
55    # Sort revs
56    rows = [[fields[i].upper() for i in print_fields] + extra_fields]
57
58    prev = [''] * len(print_fields)
59    prev_commit = None
60    commit_cnt = 0
61    commit_field = fields.index('commit')
62    for rev in revs:
63        # Only use fields that we want to print
64        cols = [rev[i] for i in print_fields]
65
66
67        if cols != prev:
68            commit_cnt = 1
69            test_run_cnt = 1
70            new_row = [''] * (len(print_fields) + len(extra_fields))
71
72            for i in print_fields:
73                if cols[i] != prev[i]:
74                    break
75            new_row[i:-len(extra_fields)] = cols[i:]
76            rows.append(new_row)
77        else:
78            if rev[commit_field] != prev_commit:
79                commit_cnt += 1
80            test_run_cnt += 1
81
82        if verbosity < 2:
83            new_row[-2] = commit_cnt
84        new_row[-1] = test_run_cnt
85        prev = cols
86        prev_commit = rev[commit_field]
87
88    print_table(rows)
89
90def is_xml_format(repo, commit):
91    """Check if the commit contains xml (or json) data"""
92    if repo.rev_parse(commit + ':results.xml'):
93        log.debug("Detected report in xml format in %s", commit)
94        return True
95    else:
96        log.debug("No xml report in %s, assuming json formatted results", commit)
97        return False
98
99def read_results(repo, tags, xml=True):
100    """Read result files from repo"""
101
102    def parse_xml_stream(data):
103        """Parse multiple concatenated XML objects"""
104        objs = []
105        xml_d = ""
106        for line in data.splitlines():
107            if xml_d and line.startswith('<?xml version='):
108                objs.append(ET.fromstring(xml_d))
109                xml_d = line
110            else:
111                xml_d += line
112        objs.append(ET.fromstring(xml_d))
113        return objs
114
115    def parse_json_stream(data):
116        """Parse multiple concatenated JSON objects"""
117        objs = []
118        json_d = ""
119        for line in data.splitlines():
120            if line == '}{':
121                json_d += '}'
122                objs.append(json.loads(json_d, object_pairs_hook=OrderedDict))
123                json_d = '{'
124            else:
125                json_d += line
126        objs.append(json.loads(json_d, object_pairs_hook=OrderedDict))
127        return objs
128
129    num_revs = len(tags)
130
131    # Optimize by reading all data with one git command
132    log.debug("Loading raw result data from %d tags, %s...", num_revs, tags[0])
133    if xml:
134        git_objs = [tag + ':metadata.xml' for tag in tags] + [tag + ':results.xml' for tag in tags]
135        data = parse_xml_stream(repo.run_cmd(['show'] + git_objs + ['--']))
136        return ([metadata_xml_to_json(e) for e in data[0:num_revs]],
137                [results_xml_to_json(e) for e in data[num_revs:]])
138    else:
139        git_objs = [tag + ':metadata.json' for tag in tags] + [tag + ':results.json' for tag in tags]
140        data = parse_json_stream(repo.run_cmd(['show'] + git_objs + ['--']))
141        return data[0:num_revs], data[num_revs:]
142
143
144def get_data_item(data, key):
145    """Nested getitem lookup"""
146    for k in key.split('.'):
147        data = data[k]
148    return data
149
150
151def metadata_diff(metadata_l, metadata_r):
152    """Prepare a metadata diff for printing"""
153    keys = [('Hostname', 'hostname', 'hostname'),
154            ('Branch', 'branch', 'layers.meta.branch'),
155            ('Commit number', 'commit_num', 'layers.meta.commit_count'),
156            ('Commit', 'commit', 'layers.meta.commit'),
157            ('Number of test runs', 'testrun_count', 'testrun_count')
158           ]
159
160    def _metadata_diff(key):
161        """Diff metadata from two test reports"""
162        try:
163            val1 = get_data_item(metadata_l, key)
164        except KeyError:
165            val1 = '(N/A)'
166        try:
167            val2 = get_data_item(metadata_r, key)
168        except KeyError:
169            val2 = '(N/A)'
170        return val1, val2
171
172    metadata = OrderedDict()
173    for title, key, key_json in keys:
174        value_l, value_r = _metadata_diff(key_json)
175        metadata[key] = {'title': title,
176                         'value_old': value_l,
177                         'value': value_r}
178    return metadata
179
180
181def print_diff_report(metadata_l, data_l, metadata_r, data_r):
182    """Print differences between two data sets"""
183
184    # First, print general metadata
185    print("\nTEST METADATA:\n==============")
186    meta_diff = metadata_diff(metadata_l, metadata_r)
187    rows = []
188    row_fmt = ['{:{wid}} ', '{:<{wid}}   ', '{:<{wid}}']
189    rows = [['', 'CURRENT COMMIT', 'COMPARING WITH']]
190    for key, val in meta_diff.items():
191        # Shorten commit hashes
192        if key == 'commit':
193            rows.append([val['title'] + ':', val['value'][:20], val['value_old'][:20]])
194        else:
195            rows.append([val['title'] + ':', val['value'], val['value_old']])
196    print_table(rows, row_fmt)
197
198
199    # Print test results
200    print("\nTEST RESULTS:\n=============")
201
202    tests = list(data_l['tests'].keys())
203    # Append tests that are only present in 'right' set
204    tests += [t for t in list(data_r['tests'].keys()) if t not in tests]
205
206    # Prepare data to be printed
207    rows = []
208    row_fmt = ['{:8}', '{:{wid}}', '{:{wid}}', '  {:>{wid}}', ' {:{wid}} ', '{:{wid}}',
209               '  {:>{wid}}', '  {:>{wid}}']
210    num_cols = len(row_fmt)
211    for test in tests:
212        test_l = data_l['tests'][test] if test in data_l['tests'] else None
213        test_r = data_r['tests'][test] if test in data_r['tests'] else None
214        pref = ' '
215        if test_l is None:
216            pref = '+'
217        elif test_r is None:
218            pref = '-'
219        descr = test_l['description'] if test_l else test_r['description']
220        heading = "{} {}: {}".format(pref, test, descr)
221
222        rows.append([heading])
223
224        # Generate the list of measurements
225        meas_l = test_l['measurements'] if test_l else {}
226        meas_r = test_r['measurements'] if test_r else {}
227        measurements = list(meas_l.keys())
228        measurements += [m for m in list(meas_r.keys()) if m not in measurements]
229
230        for meas in measurements:
231            m_pref = ' '
232            if meas in meas_l:
233                stats_l = measurement_stats(meas_l[meas], 'l.')
234            else:
235                stats_l = measurement_stats(None, 'l.')
236                m_pref = '+'
237            if meas in meas_r:
238                stats_r = measurement_stats(meas_r[meas], 'r.')
239            else:
240                stats_r = measurement_stats(None, 'r.')
241                m_pref = '-'
242            stats = stats_l.copy()
243            stats.update(stats_r)
244
245            absdiff = stats['val_cls'](stats['r.mean'] - stats['l.mean'])
246            reldiff = "{:+.1f} %".format(absdiff * 100 / stats['l.mean'])
247            if stats['r.mean'] > stats['l.mean']:
248                absdiff = '+' + str(absdiff)
249            else:
250                absdiff = str(absdiff)
251            rows.append(['', m_pref, stats['name'] + ' ' + stats['quantity'],
252                         str(stats['l.mean']), '->', str(stats['r.mean']),
253                         absdiff, reldiff])
254        rows.append([''] * num_cols)
255
256    print_table(rows, row_fmt)
257
258    print()
259
260
261class BSSummary(object):
262    def __init__(self, bs1, bs2):
263        self.tasks = {'count': bs2.num_tasks,
264                      'change': '{:+d}'.format(bs2.num_tasks - bs1.num_tasks)}
265        self.top_consumer = None
266        self.top_decrease = None
267        self.top_increase = None
268        self.ver_diff = OrderedDict()
269
270        tasks_diff = diff_buildstats(bs1, bs2, 'cputime')
271
272        # Get top consumers of resources
273        tasks_diff = sorted(tasks_diff, key=attrgetter('value2'))
274        self.top_consumer = tasks_diff[-5:]
275
276        # Get biggest increase and decrease in resource usage
277        tasks_diff = sorted(tasks_diff, key=attrgetter('absdiff'))
278        self.top_decrease = tasks_diff[0:5]
279        self.top_increase = tasks_diff[-5:]
280
281        # Compare recipe versions and prepare data for display
282        ver_diff = BSVerDiff(bs1, bs2)
283        if ver_diff:
284            if ver_diff.new:
285                self.ver_diff['New recipes'] = [(n, r.evr) for n, r in ver_diff.new.items()]
286            if ver_diff.dropped:
287                self.ver_diff['Dropped recipes'] = [(n, r.evr) for n, r in ver_diff.dropped.items()]
288            if ver_diff.echanged:
289                self.ver_diff['Epoch changed'] = [(n, "{} &rarr; {}".format(r.left.evr, r.right.evr)) for n, r in ver_diff.echanged.items()]
290            if ver_diff.vchanged:
291                self.ver_diff['Version changed'] = [(n, "{} &rarr; {}".format(r.left.version, r.right.version)) for n, r in ver_diff.vchanged.items()]
292            if ver_diff.rchanged:
293                self.ver_diff['Revision changed'] = [(n, "{} &rarr; {}".format(r.left.evr, r.right.evr)) for n, r in ver_diff.rchanged.items()]
294
295
296def print_html_report(data, id_comp, buildstats):
297    """Print report in html format"""
298    # Handle metadata
299    metadata = metadata_diff(data[id_comp].metadata, data[-1].metadata)
300
301    # Generate list of tests
302    tests = []
303    for test in data[-1].results['tests'].keys():
304        test_r = data[-1].results['tests'][test]
305        new_test = {'name': test_r['name'],
306                    'description': test_r['description'],
307                    'status': test_r['status'],
308                    'measurements': [],
309                    'err_type': test_r.get('err_type'),
310                   }
311        # Limit length of err output shown
312        if 'message' in test_r:
313            lines = test_r['message'].splitlines()
314            if len(lines) > 20:
315                new_test['message'] = '...\n' + '\n'.join(lines[-20:])
316            else:
317                new_test['message'] = test_r['message']
318
319
320        # Generate the list of measurements
321        for meas in test_r['measurements'].keys():
322            meas_r = test_r['measurements'][meas]
323            meas_type = 'time' if meas_r['type'] == 'sysres' else 'size'
324            new_meas = {'name': meas_r['name'],
325                        'legend': meas_r['legend'],
326                        'description': meas_r['name'] + ' ' + meas_type,
327                       }
328            samples = []
329
330            # Run through all revisions in our data
331            for meta, test_data in data:
332                if (not test in test_data['tests'] or
333                        not meas in test_data['tests'][test]['measurements']):
334                    samples.append(measurement_stats(None))
335                    continue
336                test_i = test_data['tests'][test]
337                meas_i = test_i['measurements'][meas]
338                commit_num = get_data_item(meta, 'layers.meta.commit_count')
339                commit = get_data_item(meta, 'layers.meta.commit')
340                # Add start_time for both test measurement types of sysres and disk usage
341                start_time = test_i['start_time'][0]
342                samples.append(measurement_stats(meas_i, '', start_time))
343                samples[-1]['commit_num'] = commit_num
344                samples[-1]['commit'] = commit
345
346            absdiff = samples[-1]['val_cls'](samples[-1]['mean'] - samples[id_comp]['mean'])
347            reldiff = absdiff * 100 / samples[id_comp]['mean']
348            new_meas['absdiff'] = absdiff
349            new_meas['absdiff_str'] = str(absdiff) if absdiff < 0 else '+' + str(absdiff)
350            new_meas['reldiff'] = reldiff
351            new_meas['reldiff_str'] = "{:+.1f} %".format(reldiff)
352            new_meas['samples'] = samples
353            new_meas['value'] = samples[-1]
354            new_meas['value_type'] = samples[-1]['val_cls']
355
356            # Compare buildstats
357            bs_key = test + '.' + meas
358            rev = str(metadata['commit_num']['value'])
359            comp_rev = str(metadata['commit_num']['value_old'])
360            if (buildstats and rev in buildstats and bs_key in buildstats[rev] and
361                    comp_rev in buildstats and bs_key in buildstats[comp_rev]):
362                new_meas['buildstats'] = BSSummary(buildstats[comp_rev][bs_key],
363                                                   buildstats[rev][bs_key])
364
365
366            new_test['measurements'].append(new_meas)
367        tests.append(new_test)
368
369    # Chart options
370    chart_opts = {'haxis': {'min': get_data_item(data[0][0], 'layers.meta.commit_count'),
371                            'max': get_data_item(data[-1][0], 'layers.meta.commit_count')}
372                 }
373
374    print(html.template.render(title="Build Perf Test Report",
375                               metadata=metadata, test_data=tests,
376                               chart_opts=chart_opts))
377
378
379def get_buildstats(repo, notes_ref, notes_ref2, revs, outdir=None):
380    """Get the buildstats from git notes"""
381    full_ref = 'refs/notes/' + notes_ref
382    if not repo.rev_parse(full_ref):
383        log.error("No buildstats found, please try running "
384                  "'git fetch origin %s:%s' to fetch them from the remote",
385                  full_ref, full_ref)
386        return
387
388    missing = False
389    buildstats = {}
390    log.info("Parsing buildstats from 'refs/notes/%s'", notes_ref)
391    for rev in revs:
392        buildstats[rev.commit_number] = {}
393        log.debug('Dumping buildstats for %s (%s)', rev.commit_number,
394                  rev.commit)
395        for tag in rev.tags:
396            log.debug('    %s', tag)
397            try:
398                try:
399                    bs_all = json.loads(repo.run_cmd(['notes', '--ref', notes_ref, 'show', tag + '^0']))
400                except GitError:
401                    if notes_ref2:
402                        bs_all = json.loads(repo.run_cmd(['notes', '--ref', notes_ref2, 'show', tag + '^0']))
403                    else:
404                        raise
405            except GitError:
406                log.warning("Buildstats not found for %s", tag)
407                bs_all = {}
408                missing = True
409
410            for measurement, bs in bs_all.items():
411                # Write out onto disk
412                if outdir:
413                    tag_base, run_id = tag.rsplit('/', 1)
414                    tag_base = tag_base.replace('/', '_')
415                    bs_dir = os.path.join(outdir, measurement, tag_base)
416                    if not os.path.exists(bs_dir):
417                        os.makedirs(bs_dir)
418                    with open(os.path.join(bs_dir, run_id + '.json'), 'w') as f:
419                        json.dump(bs, f, indent=2)
420
421                # Read buildstats into a dict
422                _bs = BuildStats.from_json(bs)
423                if measurement not in buildstats[rev.commit_number]:
424                    buildstats[rev.commit_number][measurement] = _bs
425                else:
426                    buildstats[rev.commit_number][measurement].aggregate(_bs)
427
428    if missing:
429        log.info("Buildstats were missing for some test runs, please "
430                 "run 'git fetch origin %s:%s' and try again",
431                 full_ref, full_ref)
432
433    return buildstats
434
435
436def auto_args(repo, args):
437    """Guess arguments, if not defined by the user"""
438    # Get the latest commit in the repo
439    log.debug("Guessing arguments from the latest commit")
440    msg = repo.run_cmd(['log', '-1', '--branches', '--remotes', '--format=%b'])
441    for line in msg.splitlines():
442        split = line.split(':', 1)
443        if len(split) != 2:
444            continue
445
446        key = split[0]
447        val = split[1].strip()
448        if key == 'hostname' and not args.hostname:
449            log.debug("Using hostname %s", val)
450            args.hostname = val
451        elif key == 'branch' and not args.branch:
452            log.debug("Using branch %s", val)
453            args.branch = val
454
455
456def parse_args(argv):
457    """Parse command line arguments"""
458    description = """
459Examine build performance test results from a Git repository"""
460    parser = argparse.ArgumentParser(
461        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
462        description=description)
463
464    parser.add_argument('--debug', '-d', action='store_true',
465                        help="Verbose logging")
466    parser.add_argument('--repo', '-r', required=True,
467                        help="Results repository (local git clone)")
468    parser.add_argument('--list', '-l', action='count',
469                        help="List available test runs")
470    parser.add_argument('--html', action='store_true',
471                        help="Generate report in html format")
472    group = parser.add_argument_group('Tag and revision')
473    group.add_argument('--tag-name', '-t',
474                       default='{hostname}/{branch}/{machine}/{commit_number}-g{commit}/{tag_number}',
475                       help="Tag name (pattern) for finding results")
476    group.add_argument('--hostname', '-H')
477    group.add_argument('--branch', '-B', default='master', help="Branch to find commit in")
478    group.add_argument('--branch2', help="Branch to find comparision revisions in")
479    group.add_argument('--machine', default='qemux86')
480    group.add_argument('--history-length', default=300, type=int,
481                       help="Number of tested revisions to plot in html report")
482    group.add_argument('--commit',
483                       help="Revision to search for")
484    group.add_argument('--commit-number',
485                       help="Revision number to search for, redundant if "
486                            "--commit is specified")
487    group.add_argument('--commit2',
488                       help="Revision to compare with")
489    group.add_argument('--commit-number2',
490                       help="Revision number to compare with, redundant if "
491                            "--commit2 is specified")
492    parser.add_argument('--dump-buildstats', nargs='?', const='.',
493                        help="Dump buildstats of the tests")
494
495    return parser.parse_args(argv)
496
497
498def main(argv=None):
499    """Script entry point"""
500    args = parse_args(argv)
501    if args.debug:
502        log.setLevel(logging.DEBUG)
503
504    repo = GitRepo(args.repo)
505
506    if args.list:
507        list_test_revs(repo, args.tag_name, args.list, hostname=args.hostname)
508        return 0
509
510    # Determine hostname which to use
511    if not args.hostname:
512        auto_args(repo, args)
513
514    revs = gitarchive.get_test_revs(log, repo, args.tag_name, hostname=args.hostname,
515                                    branch=args.branch, machine=args.machine)
516    if args.branch2 and args.branch2 != args.branch:
517        revs2 = gitarchive.get_test_revs(log, repo, args.tag_name, hostname=args.hostname,
518                                         branch=args.branch2, machine=args.machine)
519        if not len(revs2):
520            log.error("No revisions found to compare against")
521            return 1
522        if not len(revs):
523            log.error("No revision to report on found")
524            return 1
525    else:
526        if len(revs) < 2:
527            log.error("Only %d tester revisions found, unable to generate report" % len(revs))
528            return 1
529
530    # Pick revisions
531    if args.commit:
532        if args.commit_number:
533            log.warning("Ignoring --commit-number as --commit was specified")
534        index1 = gitarchive.rev_find(revs, 'commit', args.commit)
535    elif args.commit_number:
536        index1 = gitarchive.rev_find(revs, 'commit_number', args.commit_number)
537    else:
538        index1 = len(revs) - 1
539
540    if args.branch2 and args.branch2 != args.branch:
541        revs2.append(revs[index1])
542        index1 = len(revs2) - 1
543        revs = revs2
544
545    if args.commit2:
546        if args.commit_number2:
547            log.warning("Ignoring --commit-number2 as --commit2 was specified")
548        index2 = gitarchive.rev_find(revs, 'commit', args.commit2)
549    elif args.commit_number2:
550        index2 = gitarchive.rev_find(revs, 'commit_number', args.commit_number2)
551    else:
552        if index1 > 0:
553            index2 = index1 - 1
554            # Find the closest matching commit number for comparision
555            # In future we could check the commit is a common ancestor and
556            # continue back if not but this good enough for now
557            while index2 > 0 and revs[index2].commit_number > revs[index1].commit_number:
558                index2 = index2 - 1
559        else:
560            log.error("Unable to determine the other commit, use "
561                      "--commit2 or --commit-number2 to specify it")
562            return 1
563
564    index_l = min(index1, index2)
565    index_r = max(index1, index2)
566
567    rev_l = revs[index_l]
568    rev_r = revs[index_r]
569    log.debug("Using 'left' revision %s (%s), %s test runs:\n    %s",
570              rev_l.commit_number, rev_l.commit, len(rev_l.tags),
571              '\n    '.join(rev_l.tags))
572    log.debug("Using 'right' revision %s (%s), %s test runs:\n    %s",
573              rev_r.commit_number, rev_r.commit, len(rev_r.tags),
574              '\n    '.join(rev_r.tags))
575
576    # Check report format used in the repo (assume all reports in the same fmt)
577    xml = is_xml_format(repo, revs[index_r].tags[-1])
578
579    if args.html:
580        index_0 = max(0, min(index_l, index_r - args.history_length))
581        rev_range = range(index_0, index_r + 1)
582    else:
583        # We do not need range of commits for text report (no graphs)
584        index_0 = index_l
585        rev_range = (index_l, index_r)
586
587    # Read raw data
588    log.debug("Reading %d revisions, starting from %s (%s)",
589              len(rev_range), revs[index_0].commit_number, revs[index_0].commit)
590    raw_data = [read_results(repo, revs[i].tags, xml) for i in rev_range]
591
592    data = []
593    for raw_m, raw_d in raw_data:
594        data.append(AggregateTestData(aggregate_metadata(raw_m),
595                                      aggregate_data(raw_d)))
596
597    # Read buildstats only when needed
598    buildstats = None
599    if args.dump_buildstats or args.html:
600        outdir = 'oe-build-perf-buildstats' if args.dump_buildstats else None
601        notes_ref = 'buildstats/{}/{}/{}'.format(args.hostname, args.branch, args.machine)
602        notes_ref2 = None
603        if args.branch2:
604            notes_ref = 'buildstats/{}/{}/{}'.format(args.hostname, args.branch2, args.machine)
605            notes_ref2 = 'buildstats/{}/{}/{}'.format(args.hostname, args.branch, args.machine)
606        buildstats = get_buildstats(repo, notes_ref, notes_ref2, [rev_l, rev_r], outdir)
607
608    # Print report
609    if not args.html:
610        print_diff_report(data[0].metadata, data[0].results,
611                          data[1].metadata, data[1].results)
612    else:
613        # Re-map 'left' list index to the data table where index_0 maps to 0
614        print_html_report(data, index_l - index_0, buildstats)
615
616    return 0
617
618if __name__ == "__main__":
619    sys.exit(main())
620