1#!/usr/bin/env python 2# Copyright 2015 gRPC authors. 3# 4# Licensed under the Apache License, Version 2.0 (the "License"); 5# you may not use this file except in compliance with the License. 6# You may obtain a copy of the License at 7# 8# http://www.apache.org/licenses/LICENSE-2.0 9# 10# Unless required by applicable law or agreed to in writing, software 11# distributed under the License is distributed on an "AS IS" BASIS, 12# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13# See the License for the specific language governing permissions and 14# limitations under the License. 15"""Detect new flakes and create issues for them""" 16 17from __future__ import absolute_import 18from __future__ import division 19from __future__ import print_function 20 21import datetime 22import json 23import logging 24import os 25import pprint 26import sys 27import urllib 28import urllib2 29from collections import namedtuple 30 31gcp_utils_dir = os.path.abspath( 32 os.path.join(os.path.dirname(__file__), '../gcp/utils')) 33sys.path.append(gcp_utils_dir) 34 35import big_query_utils 36 37GH_ISSUE_CREATION_URL = 'https://api.github.com/repos/grpc/grpc/issues' 38GH_ISSUE_SEARCH_URL = 'https://api.github.com/search/issues' 39KOKORO_BASE_URL = 'https://kokoro2.corp.google.com/job/' 40 41 42def gh(url, data=None): 43 request = urllib2.Request(url, data=data) 44 assert TOKEN 45 request.add_header('Authorization', 'token {}'.format(TOKEN)) 46 if data: 47 request.add_header('Content-type', 'application/json') 48 response = urllib2.urlopen(request) 49 if 200 <= response.getcode() < 300: 50 return json.loads(response.read()) 51 else: 52 raise ValueError('Error ({}) accessing {}'.format( 53 response.getcode(), response.geturl())) 54 55 56def search_gh_issues(search_term, status='open'): 57 params = ' '.join((search_term, 'is:issue', 'is:open', 'repo:grpc/grpc')) 58 qargs = urllib.urlencode({'q': params}) 59 url = '?'.join((GH_ISSUE_SEARCH_URL, qargs)) 60 response = gh(url) 61 return response 62 63 64def create_gh_issue(title, body, labels, assignees=[]): 65 params = {'title': title, 'body': body, 'labels': labels} 66 if assignees: 67 params['assignees'] = assignees 68 data = json.dumps(params) 69 response = gh(GH_ISSUE_CREATION_URL, data) 70 issue_url = response['html_url'] 71 print('Created issue {} for {}'.format(issue_url, title)) 72 73 74def build_kokoro_url(job_name, build_id): 75 job_path = '{}/{}'.format('/job/'.join(job_name.split('/')), build_id) 76 return KOKORO_BASE_URL + job_path 77 78 79def create_issues(new_flakes, always_create): 80 for test_name, results_row in new_flakes.items(): 81 poll_strategy, job_name, build_id, timestamp = results_row 82 # TODO(dgq): the Kokoro URL has a limited lifetime. The permanent and ideal 83 # URL would be the sponge one, but there's currently no easy way to retrieve 84 # it. 85 url = build_kokoro_url(job_name, build_id) 86 title = 'New Failure: ' + test_name 87 body = '- Test: {}\n- Poll Strategy: {}\n- URL: {}'.format( 88 test_name, poll_strategy, url) 89 labels = ['infra/New Failure'] 90 if always_create: 91 proceed = True 92 else: 93 preexisting_issues = search_gh_issues(test_name) 94 if preexisting_issues['total_count'] > 0: 95 print('\nFound {} issues for "{}":'.format( 96 preexisting_issues['total_count'], test_name)) 97 for issue in preexisting_issues['items']: 98 print('\t"{}" ; URL: {}'.format(issue['title'], 99 issue['html_url'])) 100 else: 101 print( 102 '\nNo preexisting issues found for "{}"'.format(test_name)) 103 proceed = raw_input( 104 'Create issue for:\nTitle: {}\nBody: {}\n[Y/n] '.format( 105 title, body)) in ('y', 'Y', '') 106 if proceed: 107 assignees_str = raw_input( 108 'Asignees? (comma-separated, leave blank for unassigned): ') 109 assignees = [ 110 assignee.strip() for assignee in assignees_str.split(',') 111 ] 112 create_gh_issue(title, body, labels, assignees) 113 114 115def print_table(table, format): 116 first_time = True 117 for test_name, results_row in table.items(): 118 poll_strategy, job_name, build_id, timestamp = results_row 119 full_kokoro_url = build_kokoro_url(job_name, build_id) 120 if format == 'human': 121 print("\t- Test: {}, Polling: {}, Timestamp: {}, url: {}".format( 122 test_name, poll_strategy, timestamp, full_kokoro_url)) 123 else: 124 assert (format == 'csv') 125 if first_time: 126 print('test,timestamp,url') 127 first_time = False 128 print("{},{},{}".format(test_name, timestamp, full_kokoro_url)) 129 130 131Row = namedtuple('Row', ['poll_strategy', 'job_name', 'build_id', 'timestamp']) 132 133 134def get_new_failures(dates): 135 bq = big_query_utils.create_big_query() 136 this_script_path = os.path.join(os.path.dirname(__file__)) 137 sql_script = os.path.join(this_script_path, 'sql/new_failures_24h.sql') 138 with open(sql_script) as query_file: 139 query = query_file.read().format( 140 calibration_begin=dates['calibration']['begin'], 141 calibration_end=dates['calibration']['end'], 142 reporting_begin=dates['reporting']['begin'], 143 reporting_end=dates['reporting']['end']) 144 logging.debug("Query:\n%s", query) 145 query_job = big_query_utils.sync_query_job(bq, 'grpc-testing', query) 146 page = bq.jobs().getQueryResults( 147 pageToken=None, **query_job['jobReference']).execute(num_retries=3) 148 rows = page.get('rows') 149 if rows: 150 return { 151 row['f'][0]['v']: Row( 152 poll_strategy=row['f'][1]['v'], 153 job_name=row['f'][2]['v'], 154 build_id=row['f'][3]['v'], 155 timestamp=row['f'][4]['v']) 156 for row in rows 157 } 158 else: 159 return {} 160 161 162def parse_isodate(date_str): 163 return datetime.datetime.strptime(date_str, "%Y-%m-%d").date() 164 165 166def get_new_flakes(args): 167 """The from_date_str argument marks the beginning of the "calibration", used 168 to establish the set of pre-existing flakes, which extends over 169 "calibration_days". After the calibration period, "reporting_days" is the 170 length of time during which new flakes will be reported. 171 172from 173date 174 |--------------------|---------------| 175 ^____________________^_______________^ 176 calibration reporting 177 days days 178 """ 179 dates = process_date_args(args) 180 new_failures = get_new_failures(dates) 181 logging.info('|new failures| = %d', len(new_failures)) 182 return new_failures 183 184 185def build_args_parser(): 186 import argparse, datetime 187 parser = argparse.ArgumentParser() 188 today = datetime.date.today() 189 a_week_ago = today - datetime.timedelta(days=7) 190 parser.add_argument( 191 '--calibration_days', 192 type=int, 193 default=7, 194 help='How many days to consider for pre-existing flakes.') 195 parser.add_argument( 196 '--reporting_days', 197 type=int, 198 default=1, 199 help='How many days to consider for the detection of new flakes.') 200 parser.add_argument( 201 '--count_only', 202 dest='count_only', 203 action='store_true', 204 help='Display only number of new flakes.') 205 parser.set_defaults(count_only=False) 206 parser.add_argument( 207 '--create_issues', 208 dest='create_issues', 209 action='store_true', 210 help='Create issues for all new flakes.') 211 parser.set_defaults(create_issues=False) 212 parser.add_argument( 213 '--always_create_issues', 214 dest='always_create_issues', 215 action='store_true', 216 help='Always create issues for all new flakes. Otherwise,' 217 ' interactively prompt for every issue.') 218 parser.set_defaults(always_create_issues=False) 219 parser.add_argument( 220 '--token', 221 type=str, 222 default='', 223 help='GitHub token to use its API with a higher rate limit') 224 parser.add_argument( 225 '--format', 226 type=str, 227 choices=['human', 'csv'], 228 default='human', 229 help='Output format: are you a human or a machine?') 230 parser.add_argument( 231 '--loglevel', 232 type=str, 233 choices=['INFO', 'DEBUG', 'WARNING', 'ERROR', 'CRITICAL'], 234 default='WARNING', 235 help='Logging level.') 236 return parser 237 238 239def process_date_args(args): 240 calibration_begin = ( 241 datetime.date.today() - datetime.timedelta(days=args.calibration_days) - 242 datetime.timedelta(days=args.reporting_days)) 243 calibration_end = calibration_begin + datetime.timedelta( 244 days=args.calibration_days) 245 reporting_begin = calibration_end 246 reporting_end = reporting_begin + datetime.timedelta( 247 days=args.reporting_days) 248 return { 249 'calibration': { 250 'begin': calibration_begin, 251 'end': calibration_end 252 }, 253 'reporting': { 254 'begin': reporting_begin, 255 'end': reporting_end 256 } 257 } 258 259 260def main(): 261 global TOKEN 262 args_parser = build_args_parser() 263 args = args_parser.parse_args() 264 if args.create_issues and not args.token: 265 raise ValueError( 266 'Missing --token argument, needed to create GitHub issues') 267 TOKEN = args.token 268 269 logging_level = getattr(logging, args.loglevel) 270 logging.basicConfig(format='%(asctime)s %(message)s', level=logging_level) 271 new_flakes = get_new_flakes(args) 272 273 dates = process_date_args(args) 274 275 dates_info_string = 'from {} until {} (calibrated from {} until {})'.format( 276 dates['reporting']['begin'].isoformat(), 277 dates['reporting']['end'].isoformat(), 278 dates['calibration']['begin'].isoformat(), 279 dates['calibration']['end'].isoformat()) 280 281 if args.format == 'human': 282 if args.count_only: 283 print(len(new_flakes), dates_info_string) 284 elif new_flakes: 285 found_msg = 'Found {} new flakes {}'.format( 286 len(new_flakes), dates_info_string) 287 print(found_msg) 288 print('*' * len(found_msg)) 289 print_table(new_flakes, 'human') 290 if args.create_issues: 291 create_issues(new_flakes, args.always_create_issues) 292 else: 293 print('No new flakes found '.format(len(new_flakes)), 294 dates_info_string) 295 elif args.format == 'csv': 296 if args.count_only: 297 print('from_date,to_date,count') 298 print('{},{},{}'.format(dates['reporting']['begin'].isoformat(), 299 dates['reporting']['end'].isoformat(), 300 len(new_flakes))) 301 else: 302 print_table(new_flakes, 'csv') 303 else: 304 raise ValueError('Invalid argument for --format: {}'.format( 305 args.format)) 306 307 308if __name__ == '__main__': 309 main() 310