Scott Zawalski | 20a9b58 | 2011-11-21 11:49:40 -0800 | [diff] [blame] | 1 | #!/usr/bin/python |
| 2 | # |
| 3 | # Copyright (c) 2011 The Chromium OS Authors. All rights reserved. |
| 4 | # Use of this source code is governed by a BSD-style license that can be |
| 5 | # found in the LICENSE file. |
| 6 | |
| 7 | """Executes on all unlocked hosts in Autotest lab in parallel at a given rate. |
| 8 | |
| 9 | Used to run a command or script on all hosts, or only those of a given platform, |
| 10 | in the Autotest lab. Allows a configurable number of commands to be started in |
| 11 | parallel. |
| 12 | """ |
| 13 | |
| 14 | |
| 15 | import datetime |
| 16 | import logging |
| 17 | import optparse |
| 18 | import os |
| 19 | import time |
| 20 | |
| 21 | import chromeos_test_common |
| 22 | from chromeos_test import autotest_util |
| 23 | from chromeos_test import common_util |
| 24 | from chromeos_test import mp_log_util |
| 25 | from chromeos_test import mp_thread_pool as tp |
| 26 | |
| 27 | # Default number of hosts to run command/script in parallel. |
| 28 | DEFAULT_CONCURRENCY = 64 |
| 29 | |
| 30 | # Default number of hosts to update in parallel. |
| 31 | DEFAULT_UPDATE_CONCURRENCY = 24 |
| 32 | |
| 33 | # Default location of ChromeOS checkout. |
| 34 | DEFAULT_GCLIENT_ROOT = '/usr/local/google/home/${USER}/chromeos' |
| 35 | |
| 36 | # Default path for individual host logs. Each host will have it's own file. E.g. |
| 37 | # <default_log_path>/<host>.log |
| 38 | DEFAULT_LOG_PATH = ('/tmp/mass_command_logs/%s/' |
| 39 | % time.strftime('%Y-%m-%d-%H-%M', time.localtime())) |
| 40 | |
| 41 | # Default root path on remote device to copy scripts to |
| 42 | DEFAULT_REMOTE_COPY_PATH = '/tmp/' |
| 43 | |
| 44 | # Amount of seconds to wait before declaring an command/script has failed. |
| 45 | DEFAULT_TIMEOUT = 120 |
| 46 | |
| 47 | # Amount of seconds to wait before declaring an update has failed. |
| 48 | DEFAULT_UPDATE_TIMEOUT = 2400 |
| 49 | |
| 50 | |
| 51 | def ExecuteTask(failure_desc): |
| 52 | """Decorator for try/except/log pattern for reporting status and failures. |
| 53 | |
| 54 | Args: |
| 55 | failure_desc: Simple string description of task. |
| 56 | |
| 57 | Returns: |
| 58 | Decorator function to wrap a method call. |
| 59 | """ |
| 60 | |
| 61 | def DecoratorFunc(func): |
| 62 | """Function that takes the user called method as an argument.""" |
| 63 | |
| 64 | def WrappedFunc(self, *args): |
| 65 | """Function that wraps and executes user called method. |
| 66 | |
| 67 | Args: |
| 68 | self: Self object of the class method called by decorator. |
| 69 | args: Arguments to user called method. |
| 70 | |
| 71 | Returns: |
| 72 | True/False if user called method succeeded. |
| 73 | """ |
| 74 | try: |
| 75 | output = func(self, *args) |
| 76 | if output: |
| 77 | if self.output: |
| 78 | self.output += '\n' + output |
| 79 | else: |
| 80 | self.output = output |
| 81 | except common_util.ChromeOSTestError: |
| 82 | if self.logger: |
| 83 | self.logger.exception('Failed running %s %s.', self.host, |
| 84 | failure_desc) |
| 85 | self.result = failure_desc |
| 86 | return False |
| 87 | return True |
| 88 | |
| 89 | return WrappedFunc |
| 90 | return DecoratorFunc |
| 91 | |
| 92 | |
| 93 | class HostWorker(object): |
| 94 | """Responsible for ssh-test, locking, executing, and unlocking a host.""" |
| 95 | |
| 96 | def __init__(self, host, options): |
| 97 | """Create instance to perform work on a host. |
| 98 | |
| 99 | Args: |
| 100 | host: IP address of the host to connect to. |
| 101 | options: Command line options. |
| 102 | """ |
| 103 | self.host = host |
| 104 | self.options = options |
| 105 | self.result = None |
| 106 | self.output = None |
| 107 | self.logger = None |
| 108 | |
| 109 | def Execute(self, logger=None): |
| 110 | """Callback method to execute the requested action on the host. |
| 111 | |
| 112 | Usual sequence is to test connectivity by SSH-ing to the host, locking |
| 113 | the host in Autotest, running the command, then unlocking the host. |
| 114 | |
| 115 | Args: |
| 116 | logger: optional logger. |
| 117 | |
| 118 | Sets: |
| 119 | self.result to 'PASS' or failure ['SSH', 'LOCK', 'COPY', 'CMD', 'URL']. |
| 120 | self.output to standard out of command. |
| 121 | """ |
| 122 | try: |
| 123 | if logger: |
| 124 | # Store logger in self.logger so it is accessible in ExecuteTask. |
| 125 | self.logger = logger |
| 126 | logger.info('Executing for host %s', self.host) |
| 127 | |
| 128 | if not self.options.skip_ssh: |
| 129 | if not self.PingHost(): |
| 130 | return |
| 131 | |
| 132 | if self.options.lock: |
| 133 | if not self.LockUnlockHost(True): |
| 134 | return |
| 135 | |
| 136 | # Now that the host may be locked in Autotest the rest of the loop will |
| 137 | # execute in a try/finally to make sure the host is still unlocked if |
| 138 | # any of the remaining steps throw an exception. |
| 139 | try: |
| 140 | if self.options.url: |
| 141 | if not self.ImageHost(): |
| 142 | return |
| 143 | else: |
| 144 | cmd = self.options.cmd |
| 145 | if self.options.script: |
| 146 | cmd = self.options.remote_file |
| 147 | if not self.CopyToDevice(): |
| 148 | return |
| 149 | if not self.SSHCmdOnHost(cmd, self.options.extra_args): |
| 150 | return |
| 151 | finally: |
| 152 | if self.options.lock: |
| 153 | self.LockUnlockHost(False) |
| 154 | |
| 155 | self.result = 'PASS' |
| 156 | self.ProcessResult() |
| 157 | |
| 158 | finally: |
| 159 | # Loggers hold a thread lock which cannot be pickled, so it must be |
| 160 | # cleared before returning. |
| 161 | self.logger = None |
| 162 | |
| 163 | def ProcessResult(self): |
| 164 | """Dump the results to the screen and/or log file.""" |
| 165 | if self.logger: |
| 166 | msg = [self.host, ' finished with ', self.result] |
| 167 | |
| 168 | if self.options.echo_output: |
| 169 | if self.output: |
| 170 | msg += ['\nStdOut=[\n', self.output, '\n]'] |
| 171 | self.logger.info(''.join(msg)) |
| 172 | |
| 173 | if not self.options.no_log_files: |
| 174 | log = open(os.path.join(self.options.log_path, self.host + '.log'), 'w') |
| 175 | log.write(self.output) |
| 176 | log.close() |
| 177 | |
| 178 | @ExecuteTask('SSH') |
| 179 | def PingHost(self): |
| 180 | """Tests if the requested host is reachable over SSH.""" |
| 181 | msg = 'Failed to ssh to host=%s' % self.host |
| 182 | return common_util.RemoteCommand(self.host, 'root', 'true', error_msg=msg, |
| 183 | output=True) |
| 184 | |
| 185 | @ExecuteTask('CMD') |
| 186 | def SSHCmdOnHost(self, command, args=None): |
| 187 | """Executes a command on the target host using an SSH connection. |
| 188 | |
| 189 | Args: |
| 190 | command: Command to run. |
| 191 | args: Extra arguments to main command to run on the remote host. |
| 192 | |
| 193 | Returns: |
| 194 | String output from the command. |
| 195 | """ |
| 196 | cmd = '"%s %s"' % (command, args) |
| 197 | msg = 'Failed to run command=%s' % cmd |
| 198 | return common_util.RemoteCommand(self.host, 'root', cmd, error_msg=msg, |
| 199 | output=True) |
| 200 | |
| 201 | @ExecuteTask('COPY') |
| 202 | def CopyToDevice(self): |
| 203 | """Copies a file (usually a script file) to a host using scp. |
| 204 | |
| 205 | Returns: |
| 206 | String output from the command. |
| 207 | """ |
| 208 | msg = 'Failed to copy %s to root@%s:%s'% (self.options.script, self.host, |
| 209 | self.options.remote_file) |
| 210 | return common_util.RemoteCopy(self.host, 'root', self.options.script, |
| 211 | self.options.remote_file, error_msg=msg, |
| 212 | output=True) |
| 213 | |
| 214 | @ExecuteTask('URL') |
| 215 | def ImageHost(self): |
| 216 | """Uses the image_to_live script to update a host. |
| 217 | |
| 218 | Returns: |
| 219 | String output from the command. |
| 220 | """ |
| 221 | cmd = ('/usr/local/scripts/alarm %d %s/src/scripts/image_to_live.sh ' |
| 222 | '--update_url %s --remote %s' % (self.options.timeout, |
| 223 | self.options.gclient, |
| 224 | self.options.url, self.host)) |
| 225 | return common_util.RunCommand(cmd, output=True) |
| 226 | |
| 227 | @ExecuteTask('LOCK') |
| 228 | def LockUnlockHost(self, lock=True): |
| 229 | """Locks a host using the atest CLI. |
| 230 | |
| 231 | Locking a host tells Autotest that the host shouldn't be scheduled for |
| 232 | any other tasks. Returns true if the locking process was successful. |
| 233 | |
| 234 | Args: |
| 235 | lock: True=lock the host, False=unlock the host. |
| 236 | |
| 237 | Returns: |
| 238 | String output from the command. |
| 239 | """ |
| 240 | if lock: |
| 241 | cmd = '%s host mod -l %s' % (self.options.cli, self.host) |
| 242 | else: |
| 243 | cmd = '%s host mod -u %s' % (self.options.cli, self.host) |
| 244 | return common_util.RunCommand(cmd, output=True) |
| 245 | |
| 246 | |
| 247 | class CommandManager(object): |
| 248 | """Executes a command on all of the selected remote hosts. |
| 249 | |
| 250 | The hosts are selected from Autotest using the parameters supplied on the |
| 251 | command line. |
| 252 | """ |
| 253 | |
| 254 | def __init__(self): |
| 255 | self.options = self.ParseOptions() |
| 256 | mp_log_util.InitializeLogging(**vars(self.options)) |
| 257 | if self.options.ip_addr: |
| 258 | self.host_list = [self.options.ip_addr] |
| 259 | else: |
| 260 | self.host_list = autotest_util.GetHostList(self.options.cli, |
| 261 | self.options.acl, |
| 262 | self.options.label, |
| 263 | self.options.user, |
| 264 | self.options.status) |
| 265 | |
| 266 | @staticmethod |
| 267 | def ParseOptions(): |
| 268 | """Grab the options from the command line.""" |
| 269 | |
| 270 | parser = optparse.OptionParser( |
| 271 | 'Used to run a command or script or update on all hosts, or only those ' |
| 272 | 'of a given platform, in the Autotest lab. Allows a configurable ' |
| 273 | 'number of commands to be started in parallel.\n\n' |
| 274 | '\texample: %prog [options] command\n\n' |
| 275 | 'Arguments after command are interpreted as arguments to the command.\n' |
| 276 | '\n\texample: %prog [options] command [cmd_arg_1] [cmd_arg_2]\n\n' |
| 277 | 'Multiple command can be run by enclosing them in quotation marks.\n\n' |
| 278 | '\texample: %prog [options] "command1; command2; command2"\n\n' |
| 279 | 'When using the --script option, additional arguments are interpreted ' |
| 280 | 'as script options and are passed to the script after being copied to ' |
| 281 | 'the remote device.\n\n' |
| 282 | '\texample: %prog [options] --script /path/to/script.sh ' |
| 283 | '[script_arg_1] [script_arg_2] [script_arg_3]\n\n' |
| 284 | 'When using the --url option specify the path to the new build. ' |
| 285 | 'Additional arguments are ignored.\n\n' |
| 286 | '\texample: %prog [options] --url /path/to/build') |
| 287 | |
| 288 | # Args for describing the environment of the server machine |
| 289 | group = optparse.OptionGroup( |
| 290 | parser, 'Server Configuration', 'Options that specify the layout of ' |
| 291 | 'the machine hosting this script.') |
| 292 | group.add_option( |
| 293 | '-g', '--gclient', default=DEFAULT_GCLIENT_ROOT, |
| 294 | help=('Location of ChromeOS checkout. [default: %default]')) |
| 295 | parser.add_option_group(group) |
| 296 | |
| 297 | # Args for configuring logging. |
| 298 | group = mp_log_util.AddOptions(parser) |
| 299 | group.add_option( |
| 300 | '--log_path', default=DEFAULT_LOG_PATH, |
| 301 | help=('Where to put individual host log files. [default: %default]')) |
| 302 | group.add_option( |
| 303 | '-n', '--no_log_files', default=False, action='store_true', |
| 304 | help=('Skip writing output to files, instead display results on the ' |
| 305 | 'console window only. [default: %default]')) |
| 306 | group.add_option( |
| 307 | '-e', '--echo_output', default=False, action='store_true', |
| 308 | help=('Write command output to console. [default: %default]')) |
| 309 | parser.add_option_group(group) |
| 310 | |
| 311 | # Args for selecting machines from Autotest |
| 312 | group = autotest_util.AddOptions(parser) |
| 313 | group.add_option( |
| 314 | '-i', '--ip_addr', |
| 315 | help=('IP address of single machine to run on.')) |
| 316 | parser.add_option_group(group) |
| 317 | |
| 318 | # Args for defining how to run tasks from the server |
| 319 | group = optparse.OptionGroup( |
| 320 | parser, 'Execution Options', 'Options that define how commands are ' |
| 321 | 'run on the remote machines.') |
| 322 | group.add_option( |
| 323 | '-p', '--parallel', type='int', default=DEFAULT_CONCURRENCY, |
| 324 | help=('Number of hosts to be run concurrently. ' |
| 325 | '[default: %default].')) |
| 326 | group.add_option( |
| 327 | '-t', '--timeout', type='int', default=DEFAULT_TIMEOUT, |
| 328 | help=('Time to wait before killing the attempt to run command. ' |
| 329 | '[default: %default]')) |
| 330 | group.add_option( |
| 331 | '--skip_ssh', default=False, action='store_true', |
| 332 | help=('Skip SSH check before running on each device. ' |
| 333 | '[default: %default]')) |
| 334 | group.add_option( |
| 335 | '-l', '--lock', default=False, action='store_true', |
| 336 | help='Lock device in Autotest while running. [default: %default]') |
| 337 | parser.add_option_group(group) |
| 338 | |
| 339 | # Args for the action to take on each remote device |
| 340 | group = optparse.OptionGroup( |
| 341 | parser, 'Main Options', 'Options that define main action. Selecting ' |
| 342 | 'neither --script nor --url defaults to running a command on the ' |
| 343 | 'hosts.') |
| 344 | group.add_option( |
| 345 | '-s', '--script', nargs=2, |
| 346 | help=('Path to script to copy to host then execute. 2 args are ' |
| 347 | 'required. If the script does not take any args pass an empty ' |
| 348 | 'string \" \"')) |
| 349 | group.add_option( |
| 350 | '--url', |
| 351 | help=('Run image_to_live.sh with provided image URL. Note: Resets ' |
| 352 | 'defaults for --lock=TRUE and --timeout=2400 and --parallel=' |
| 353 | '24.')) |
| 354 | parser.add_option_group(group) |
| 355 | |
| 356 | options, args = parser.parse_args() |
| 357 | |
| 358 | options.cmd = None |
| 359 | options.extra_args = None |
| 360 | options.remote_file = None |
| 361 | |
| 362 | # If script/url was not specified, the remaining args are commands. |
| 363 | if not options.script and not options.url: |
| 364 | if not args: |
| 365 | parser.error('Either script, command, or URL must be selected.') |
| 366 | else: |
| 367 | options.cmd, options.extra_args = args[0], ' '.join(args[1:]) |
| 368 | |
| 369 | # Grab the arguments to the script and setup any extra args. |
| 370 | if options.script: |
| 371 | options.script, options.extra_args = options.script[0], options.script[1] |
| 372 | options.remote_file = os.path.join(DEFAULT_REMOTE_COPY_PATH, |
| 373 | options.script.split(os.path.sep)[-1]) |
| 374 | else: |
| 375 | options.remote_file = '' |
| 376 | |
| 377 | # For updates reset default lock and timeout. |
| 378 | if options.url: |
| 379 | # Only modify these options if they still have their default values. If |
| 380 | # the user has already overwritten them keep the users values. |
| 381 | if options.timeout == DEFAULT_TIMEOUT: |
| 382 | options.timeout = DEFAULT_UPDATE_TIMEOUT |
| 383 | if options.parallel == DEFAULT_CONCURRENCY: |
| 384 | options.parallel = DEFAULT_UPDATE_CONCURRENCY |
| 385 | |
| 386 | # Create log folder if it doesn't exist. |
| 387 | if not options.no_log_files and not os.path.exists(options.log_path): |
| 388 | os.makedirs(options.log_path) |
| 389 | |
| 390 | return options |
| 391 | |
| 392 | |
| 393 | def ProcessResults(results, result_type): |
| 394 | """Dump the results to the screen and/or log file. |
| 395 | |
| 396 | Args: |
| 397 | results: Hosts with the same result type. |
| 398 | result_type: String description of the result type. |
| 399 | """ |
| 400 | msg = '%d hosts %s.\n' % (len(results), result_type) |
| 401 | msg += ', '.join(results) |
| 402 | mp_log_util.LogWithHeader(msg, width=80, symbol='-') |
| 403 | |
| 404 | |
| 405 | def main(): |
| 406 | """Run commands in parallel on remote hosts.""" |
| 407 | script_start_time = datetime.datetime.now() |
| 408 | cm = CommandManager() |
| 409 | if not cm.host_list: |
| 410 | logging.error('No hosts found.') |
| 411 | return |
| 412 | logging.info('Found %d hosts.', len(cm.host_list)) |
| 413 | |
| 414 | # Create work object for each host retrieved. |
| 415 | hosts = [HostWorker(host, cm.options) for host in cm.host_list] |
| 416 | |
| 417 | # Submit work to pool. |
| 418 | mp_tp = tp.MultiProcWorkPool(max_threads=cm.options.parallel) |
| 419 | hosts = mp_tp.ExecuteWorkItems( |
| 420 | hosts, provide_logger=True, |
| 421 | logger_init_callback=mp_log_util.InitializeLogging, **vars(cm.options)) |
| 422 | |
| 423 | # Now that work is done, output results. |
| 424 | status_strings = {'PASS': 'succeeded', |
| 425 | 'SSH': 'failed connecting via SSH', |
| 426 | 'LOCK': 'failed locking in Autotest', |
| 427 | 'COPY': 'failed copying script', |
| 428 | 'CMD': 'failed executing command', |
| 429 | 'URL': 'failed updating image'} |
| 430 | results = {} |
| 431 | for key in status_strings: |
| 432 | results[key] = [] |
| 433 | |
| 434 | # Divide results by result type for prettier reporting. |
| 435 | for h in hosts: |
| 436 | results[h.result].append(h.host) |
| 437 | |
| 438 | # Output final results. |
| 439 | for result, hosts in results.items(): |
| 440 | if hosts: |
| 441 | ProcessResults(hosts, status_strings[result]) |
| 442 | |
| 443 | if not cm.options.no_log_files: |
| 444 | logging.info('Log files located in %s', cm.options.log_path) |
| 445 | |
| 446 | # Follow up with some timing info. |
| 447 | script_runtime = datetime.datetime.now() - script_start_time |
| 448 | logging.info('Running Time = %d.%d seconds.', |
| 449 | script_runtime.seconds, script_runtime.microseconds) |
| 450 | |
| 451 | |
| 452 | if __name__ == '__main__': |
| 453 | main() |