date: 2026-11-05T21:55:44.021427+00:00 commit: 07965a2fb67b57f6031fed70a9a3ae963571269f
$ python3 <<'PY'
import collections
import csv
import json
import subprocess
from pathlib import Path

root = Path('evidence/track1-sqlalchemy')
images = {name: subprocess.check_output(['docker', 'image', 'inspect', tag, '--format', '{{.Id}}'], text=True).strip() for name, tag in [('before', 'mtusi-messenger:5c8f605cd3bd40245109348463d3e7111379b2d3'), ('after', 'mtusi-messenger:mtusi2026lab3-track5-20261105')]}
images['series-baseline'] = next(line.strip() for line in (root / 'baseline-run.txt').read_text().splitlines() if line.startswith('sha256:'))
assert images['before'] == next(line.strip() for line in (root / 'comparison-run.txt').read_text().splitlines() if line.startswith('sha256:'))
expected = {'series-a': {'1x4', '2x2', '4x1'}, 'series-b': {'1x1', '2x1', '8x1'}, 'compare': {'8x1'}, 'control-pool': {'8x1'}, 'control-cpus': {'4x1', '8x0.5'}}
total = 0
all_rows = []
for series, configs in expected.items():
    path = root / series / 'summary.csv'
    rows = list(csv.DictReader(line for line in path.open() if not line.startswith('#')))
    assert {r['config'] for r in rows} == configs, series
    groups = collections.defaultdict(list)
    for row in rows:
        groups[row['variant'], row['config'], row['rate']].append(row)
        assert row['disturbed'] == 'False' and row['k6_exit'] == '0', row['raw_dir']
        folder = path.parent / row['raw_dir']
        data = json.loads((folder / 'k6/summary.json').read_text())
        assert 'setup_data' not in data
        metrics = data['metrics']
        count = metrics['http_reqs{scenario:measure}']['values']['count']
        failed = metrics['http_req_failed{scenario:measure}']['values']['rate']
        dropped = metrics.get('dropped_iterations{scenario:measure}', {}).get('values', {}).get('count', 0)
        latency = metrics['http_req_duration{scenario:measure}']['values']['p(95)']
        assert int(row['requests']) == count
        assert abs(float(row['ok_rps']) - count * (1-failed) / 30) <= 0.006
        assert int(row['dropped_iterations']) == dropped
        assert abs(float(row['p95_ms']) - latency) <= 0.006
        saturated = round(failed, 5) > .01 or dropped > .01 * int(row['rate']) * 30 or round(latency, 2) > 500
        assert (row['saturated'] == 'True') == saturated
        assert all(float(row['cpu_'+name]) > 0 for name in ('vm', 'api', 'postgres', 'k6'))
        metadata = json.loads((folder / 'metadata.json').read_text())
        apis = [c for c in metadata['containers'] if '-api-' in c['name']]
        assert len(apis) == int(row['replicas'])
        expected_image = images['series-baseline' if series in ('series-a', 'series-b') else 'after' if row['variant'] == 'after' else 'before']
        assert {c['image'] for c in apis} == {expected_image}, row['raw_dir']
        assert {c['NanoCpus'] for c in apis} == {round(float(row['cpus'])*1e9)}
        assert {c['Memory'] for c in apis} == {int(row['memory_mb'])*2**20}
        assert {c['CpusetCpus'] for c in apis} == {row['api_cpuset']}
        for filename in ('docker_stats.csv', 'vm_stat.csv', 'pg_activity.csv', 'lb_stats.csv', 'pg_statements.csv'):
            samples = list(csv.DictReader(line for line in (folder / filename).open() if not line.startswith('#')))
            assert len(samples) >= 2, (folder, filename)
    assert all(len(group) == 3 and {r['repeat'] for r in group} == {'1','2','3'} for group in groups.values())
    all_rows.extend(rows)
    total += len(rows)
    print(series, len(rows), 'PASS: raw metrics, 3 repeats, actual CPU/RAM/cpuset/image, telemetry')
print('TOTAL', total, 'accepted samples; image IDs', images)

for field in ('db_mb','messages','cpu_k6','cpu_other','gap_s','wall_s','mem_api_mb','failed_ratio'):
    values = [float(r[field]) for r in all_rows]
    print(field, 'range', min(values), max(values))
print('k6 measured requests:', sum(int(r['requests']) for r in all_rows))
for path in root.glob('*/disturbed.csv'):
    with path.open() as handle:
        rejected = list(csv.DictReader(line for line in handle if not line.startswith('#')))
    print(path, 'disturbed attempts:', len(rejected))
PY

series-a 54 PASS: raw metrics, 3 repeats, actual CPU/RAM/cpuset/image, telemetry
series-b 36 PASS: raw metrics, 3 repeats, actual CPU/RAM/cpuset/image, telemetry
compare 18 PASS: raw metrics, 3 repeats, actual CPU/RAM/cpuset/image, telemetry
control-pool 15 PASS: raw metrics, 3 repeats, actual CPU/RAM/cpuset/image, telemetry
control-cpus 12 PASS: raw metrics, 3 repeats, actual CPU/RAM/cpuset/image, telemetry
TOTAL 135 accepted samples; image IDs {'before': 'sha256:3cd2f31af1b025c67f7d28931936c6f80774bd15dfea7e469e35b79cd0454388', 'after': 'sha256:676dccadf49e62f96e2692b26a2ae47848b2a2dfbb53d8753abf1635abb2b00f', 'series-baseline': 'sha256:0956b82841efe88f334e84b878a3fd746513d06a474b82ae095e6433614fd24a'}
db_mb range 33.2 4597.3
messages range 53496.0 9695935.0
cpu_k6 range 6.77 79.0
cpu_other range 1.6 71.0
gap_s range 4.0 5.6
wall_s range 41.2 42.6
mem_api_mb range 155.4 474.4
failed_ratio range 0.0 0.00016
k6 measured requests: 8271783
evidence/track1-sqlalchemy/series-b/disturbed.csv disturbed attempts: 0
evidence/track1-sqlalchemy/compare/disturbed.csv disturbed attempts: 0
evidence/track1-sqlalchemy/control-pool/disturbed.csv disturbed attempts: 0
evidence/track1-sqlalchemy/control-cpus/disturbed.csv disturbed attempts: 0
evidence/track1-sqlalchemy/series-a/disturbed.csv disturbed attempts: 0
exit: 0
