OLD | NEW |
(Empty) | |
| 1 # Copyright (c) 2006-2012 Mitch Garnaat http://garnaat.org/ |
| 2 # Copyright (c) 2012 Amazon.com, Inc. or its affiliates. |
| 3 # Copyright (c) 2010 Google |
| 4 # Copyright (c) 2008 rPath, Inc. |
| 5 # Copyright (c) 2009 The Echo Nest Corporation |
| 6 # Copyright (c) 2010, Eucalyptus Systems, Inc. |
| 7 # Copyright (c) 2011, Nexenta Systems Inc. |
| 8 # All rights reserved. |
| 9 # |
| 10 # Permission is hereby granted, free of charge, to any person obtaining a |
| 11 # copy of this software and associated documentation files (the |
| 12 # "Software"), to deal in the Software without restriction, including |
| 13 # without limitation the rights to use, copy, modify, merge, publish, dis- |
| 14 # tribute, sublicense, and/or sell copies of the Software, and to permit |
| 15 # persons to whom the Software is furnished to do so, subject to the fol- |
| 16 # lowing conditions: |
| 17 # |
| 18 # The above copyright notice and this permission notice shall be included |
| 19 # in all copies or substantial portions of the Software. |
| 20 # |
| 21 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
| 22 # OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- |
| 23 # ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT |
| 24 # SHALL THE AUTHOR BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
| 25 # WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 26 # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS |
| 27 # IN THE SOFTWARE. |
| 28 |
| 29 # |
| 30 # Parts of this code were copied or derived from sample code supplied by AWS. |
| 31 # The following notice applies to that code. |
| 32 # |
| 33 # This software code is made available "AS IS" without warranties of any |
| 34 # kind. You may copy, display, modify and redistribute the software |
| 35 # code either by itself or as incorporated into your code; provided that |
| 36 # you do not remove any proprietary notices. Your use of this software |
| 37 # code is at your own risk and you waive any claim against Amazon |
| 38 # Digital Services, Inc. or its affiliates with respect to your use of |
| 39 # this software code. (c) 2006 Amazon Digital Services, Inc. or its |
| 40 # affiliates. |
| 41 |
| 42 """ |
| 43 Handles basic connections to AWS |
| 44 """ |
| 45 |
| 46 from __future__ import with_statement |
| 47 import base64 |
| 48 import errno |
| 49 import httplib |
| 50 import os |
| 51 import Queue |
| 52 import random |
| 53 import re |
| 54 import socket |
| 55 import sys |
| 56 import time |
| 57 import urllib |
| 58 import urlparse |
| 59 import xml.sax |
| 60 import copy |
| 61 |
| 62 import auth |
| 63 import auth_handler |
| 64 import boto |
| 65 import boto.utils |
| 66 import boto.handler |
| 67 import boto.cacerts |
| 68 |
| 69 from boto import config, UserAgent |
| 70 from boto.exception import AWSConnectionError, BotoClientError |
| 71 from boto.exception import BotoServerError |
| 72 from boto.provider import Provider |
| 73 from boto.resultset import ResultSet |
| 74 |
| 75 HAVE_HTTPS_CONNECTION = False |
| 76 try: |
| 77 import ssl |
| 78 from boto import https_connection |
| 79 # Google App Engine runs on Python 2.5 so doesn't have ssl.SSLError. |
| 80 if hasattr(ssl, 'SSLError'): |
| 81 HAVE_HTTPS_CONNECTION = True |
| 82 except ImportError: |
| 83 pass |
| 84 |
| 85 try: |
| 86 import threading |
| 87 except ImportError: |
| 88 import dummy_threading as threading |
| 89 |
| 90 ON_APP_ENGINE = all(key in os.environ for key in ( |
| 91 'USER_IS_ADMIN', 'CURRENT_VERSION_ID', 'APPLICATION_ID')) |
| 92 |
| 93 PORTS_BY_SECURITY = {True: 443, |
| 94 False: 80} |
| 95 |
| 96 DEFAULT_CA_CERTS_FILE = os.path.join(os.path.dirname(os.path.abspath(boto.cacert
s.__file__ )), "cacerts.txt") |
| 97 |
| 98 |
| 99 class HostConnectionPool(object): |
| 100 |
| 101 """ |
| 102 A pool of connections for one remote (host,is_secure). |
| 103 |
| 104 When connections are added to the pool, they are put into a |
| 105 pending queue. The _mexe method returns connections to the pool |
| 106 before the response body has been read, so they connections aren't |
| 107 ready to send another request yet. They stay in the pending queue |
| 108 until they are ready for another request, at which point they are |
| 109 returned to the pool of ready connections. |
| 110 |
| 111 The pool of ready connections is an ordered list of |
| 112 (connection,time) pairs, where the time is the time the connection |
| 113 was returned from _mexe. After a certain period of time, |
| 114 connections are considered stale, and discarded rather than being |
| 115 reused. This saves having to wait for the connection to time out |
| 116 if AWS has decided to close it on the other end because of |
| 117 inactivity. |
| 118 |
| 119 Thread Safety: |
| 120 |
| 121 This class is used only fram ConnectionPool while it's mutex |
| 122 is held. |
| 123 """ |
| 124 |
| 125 def __init__(self): |
| 126 self.queue = [] |
| 127 |
| 128 def size(self): |
| 129 """ |
| 130 Returns the number of connections in the pool for this host. |
| 131 Some of the connections may still be in use, and may not be |
| 132 ready to be returned by get(). |
| 133 """ |
| 134 return len(self.queue) |
| 135 |
| 136 def put(self, conn): |
| 137 """ |
| 138 Adds a connection to the pool, along with the time it was |
| 139 added. |
| 140 """ |
| 141 self.queue.append((conn, time.time())) |
| 142 |
| 143 def get(self): |
| 144 """ |
| 145 Returns the next connection in this pool that is ready to be |
| 146 reused. Returns None of there aren't any. |
| 147 """ |
| 148 # Discard ready connections that are too old. |
| 149 self.clean() |
| 150 |
| 151 # Return the first connection that is ready, and remove it |
| 152 # from the queue. Connections that aren't ready are returned |
| 153 # to the end of the queue with an updated time, on the |
| 154 # assumption that somebody is actively reading the response. |
| 155 for _ in range(len(self.queue)): |
| 156 (conn, _) = self.queue.pop(0) |
| 157 if self._conn_ready(conn): |
| 158 return conn |
| 159 else: |
| 160 self.put(conn) |
| 161 return None |
| 162 |
| 163 def _conn_ready(self, conn): |
| 164 """ |
| 165 There is a nice state diagram at the top of httplib.py. It |
| 166 indicates that once the response headers have been read (which |
| 167 _mexe does before adding the connection to the pool), a |
| 168 response is attached to the connection, and it stays there |
| 169 until it's done reading. This isn't entirely true: even after |
| 170 the client is done reading, the response may be closed, but |
| 171 not removed from the connection yet. |
| 172 |
| 173 This is ugly, reading a private instance variable, but the |
| 174 state we care about isn't available in any public methods. |
| 175 """ |
| 176 if ON_APP_ENGINE: |
| 177 # Google AppEngine implementation of HTTPConnection doesn't contain |
| 178 # _HTTPConnection__response attribute. Moreover, it's not possible |
| 179 # to determine if given connection is ready. Reusing connections |
| 180 # simply doesn't make sense with App Engine urlfetch service. |
| 181 return False |
| 182 else: |
| 183 response = getattr(conn, '_HTTPConnection__response', None) |
| 184 return (response is None) or response.isclosed() |
| 185 |
| 186 def clean(self): |
| 187 """ |
| 188 Get rid of stale connections. |
| 189 """ |
| 190 # Note that we do not close the connection here -- somebody |
| 191 # may still be reading from it. |
| 192 while len(self.queue) > 0 and self._pair_stale(self.queue[0]): |
| 193 self.queue.pop(0) |
| 194 |
| 195 def _pair_stale(self, pair): |
| 196 """ |
| 197 Returns true of the (connection,time) pair is too old to be |
| 198 used. |
| 199 """ |
| 200 (_conn, return_time) = pair |
| 201 now = time.time() |
| 202 return return_time + ConnectionPool.STALE_DURATION < now |
| 203 |
| 204 |
| 205 class ConnectionPool(object): |
| 206 |
| 207 """ |
| 208 A connection pool that expires connections after a fixed period of |
| 209 time. This saves time spent waiting for a connection that AWS has |
| 210 timed out on the other end. |
| 211 |
| 212 This class is thread-safe. |
| 213 """ |
| 214 |
| 215 # |
| 216 # The amout of time between calls to clean. |
| 217 # |
| 218 |
| 219 CLEAN_INTERVAL = 5.0 |
| 220 |
| 221 # |
| 222 # How long before a connection becomes "stale" and won't be reused |
| 223 # again. The intention is that this time is less that the timeout |
| 224 # period that AWS uses, so we'll never try to reuse a connection |
| 225 # and find that AWS is timing it out. |
| 226 # |
| 227 # Experimentation in July 2011 shows that AWS starts timing things |
| 228 # out after three minutes. The 60 seconds here is conservative so |
| 229 # we should never hit that 3-minute timout. |
| 230 # |
| 231 |
| 232 STALE_DURATION = 60.0 |
| 233 |
| 234 def __init__(self): |
| 235 # Mapping from (host,is_secure) to HostConnectionPool. |
| 236 # If a pool becomes empty, it is removed. |
| 237 self.host_to_pool = {} |
| 238 # The last time the pool was cleaned. |
| 239 self.last_clean_time = 0.0 |
| 240 self.mutex = threading.Lock() |
| 241 ConnectionPool.STALE_DURATION = \ |
| 242 config.getfloat('Boto', 'connection_stale_duration', |
| 243 ConnectionPool.STALE_DURATION) |
| 244 |
| 245 def __getstate__(self): |
| 246 pickled_dict = copy.copy(self.__dict__) |
| 247 pickled_dict['host_to_pool'] = {} |
| 248 del pickled_dict['mutex'] |
| 249 return pickled_dict |
| 250 |
| 251 def __setstate__(self, dct): |
| 252 self.__init__() |
| 253 |
| 254 def size(self): |
| 255 """ |
| 256 Returns the number of connections in the pool. |
| 257 """ |
| 258 return sum(pool.size() for pool in self.host_to_pool.values()) |
| 259 |
| 260 def get_http_connection(self, host, is_secure): |
| 261 """ |
| 262 Gets a connection from the pool for the named host. Returns |
| 263 None if there is no connection that can be reused. It's the caller's |
| 264 responsibility to call close() on the connection when it's no longer |
| 265 needed. |
| 266 """ |
| 267 self.clean() |
| 268 with self.mutex: |
| 269 key = (host, is_secure) |
| 270 if key not in self.host_to_pool: |
| 271 return None |
| 272 return self.host_to_pool[key].get() |
| 273 |
| 274 def put_http_connection(self, host, is_secure, conn): |
| 275 """ |
| 276 Adds a connection to the pool of connections that can be |
| 277 reused for the named host. |
| 278 """ |
| 279 with self.mutex: |
| 280 key = (host, is_secure) |
| 281 if key not in self.host_to_pool: |
| 282 self.host_to_pool[key] = HostConnectionPool() |
| 283 self.host_to_pool[key].put(conn) |
| 284 |
| 285 def clean(self): |
| 286 """ |
| 287 Clean up the stale connections in all of the pools, and then |
| 288 get rid of empty pools. Pools clean themselves every time a |
| 289 connection is fetched; this cleaning takes care of pools that |
| 290 aren't being used any more, so nothing is being gotten from |
| 291 them. |
| 292 """ |
| 293 with self.mutex: |
| 294 now = time.time() |
| 295 if self.last_clean_time + self.CLEAN_INTERVAL < now: |
| 296 to_remove = [] |
| 297 for (host, pool) in self.host_to_pool.items(): |
| 298 pool.clean() |
| 299 if pool.size() == 0: |
| 300 to_remove.append(host) |
| 301 for host in to_remove: |
| 302 del self.host_to_pool[host] |
| 303 self.last_clean_time = now |
| 304 |
| 305 |
| 306 class HTTPRequest(object): |
| 307 |
| 308 def __init__(self, method, protocol, host, port, path, auth_path, |
| 309 params, headers, body): |
| 310 """Represents an HTTP request. |
| 311 |
| 312 :type method: string |
| 313 :param method: The HTTP method name, 'GET', 'POST', 'PUT' etc. |
| 314 |
| 315 :type protocol: string |
| 316 :param protocol: The http protocol used, 'http' or 'https'. |
| 317 |
| 318 :type host: string |
| 319 :param host: Host to which the request is addressed. eg. abc.com |
| 320 |
| 321 :type port: int |
| 322 :param port: port on which the request is being sent. Zero means unset, |
| 323 in which case default port will be chosen. |
| 324 |
| 325 :type path: string |
| 326 :param path: URL path that is being accessed. |
| 327 |
| 328 :type auth_path: string |
| 329 :param path: The part of the URL path used when creating the |
| 330 authentication string. |
| 331 |
| 332 :type params: dict |
| 333 :param params: HTTP url query parameters, with key as name of |
| 334 the param, and value as value of param. |
| 335 |
| 336 :type headers: dict |
| 337 :param headers: HTTP headers, with key as name of the header and value |
| 338 as value of header. |
| 339 |
| 340 :type body: string |
| 341 :param body: Body of the HTTP request. If not present, will be None or |
| 342 empty string (''). |
| 343 """ |
| 344 self.method = method |
| 345 self.protocol = protocol |
| 346 self.host = host |
| 347 self.port = port |
| 348 self.path = path |
| 349 if auth_path is None: |
| 350 auth_path = path |
| 351 self.auth_path = auth_path |
| 352 self.params = params |
| 353 # chunked Transfer-Encoding should act only on PUT request. |
| 354 if headers and 'Transfer-Encoding' in headers and \ |
| 355 headers['Transfer-Encoding'] == 'chunked' and \ |
| 356 self.method != 'PUT': |
| 357 self.headers = headers.copy() |
| 358 del self.headers['Transfer-Encoding'] |
| 359 else: |
| 360 self.headers = headers |
| 361 self.body = body |
| 362 |
| 363 def __str__(self): |
| 364 return (('method:(%s) protocol:(%s) host(%s) port(%s) path(%s) ' |
| 365 'params(%s) headers(%s) body(%s)') % (self.method, |
| 366 self.protocol, self.host, self.port, self.path, self.params, |
| 367 self.headers, self.body)) |
| 368 |
| 369 def authorize(self, connection, **kwargs): |
| 370 for key in self.headers: |
| 371 val = self.headers[key] |
| 372 if isinstance(val, unicode): |
| 373 self.headers[key] = urllib.quote_plus(val.encode('utf-8')) |
| 374 |
| 375 connection._auth_handler.add_auth(self, **kwargs) |
| 376 |
| 377 self.headers['User-Agent'] = UserAgent |
| 378 # I'm not sure if this is still needed, now that add_auth is |
| 379 # setting the content-length for POST requests. |
| 380 if 'Content-Length' not in self.headers: |
| 381 if 'Transfer-Encoding' not in self.headers or \ |
| 382 self.headers['Transfer-Encoding'] != 'chunked': |
| 383 self.headers['Content-Length'] = str(len(self.body)) |
| 384 |
| 385 |
| 386 class HTTPResponse(httplib.HTTPResponse): |
| 387 |
| 388 def __init__(self, *args, **kwargs): |
| 389 httplib.HTTPResponse.__init__(self, *args, **kwargs) |
| 390 self._cached_response = '' |
| 391 |
| 392 def read(self, amt=None): |
| 393 """Read the response. |
| 394 |
| 395 This method does not have the same behavior as |
| 396 httplib.HTTPResponse.read. Instead, if this method is called with |
| 397 no ``amt`` arg, then the response body will be cached. Subsequent |
| 398 calls to ``read()`` with no args **will return the cached response**. |
| 399 |
| 400 """ |
| 401 if amt is None: |
| 402 # The reason for doing this is that many places in boto call |
| 403 # response.read() and except to get the response body that they |
| 404 # can then process. To make sure this always works as they expect |
| 405 # we're caching the response so that multiple calls to read() |
| 406 # will return the full body. Note that this behavior only |
| 407 # happens if the amt arg is not specified. |
| 408 if not self._cached_response: |
| 409 self._cached_response = httplib.HTTPResponse.read(self) |
| 410 return self._cached_response |
| 411 else: |
| 412 return httplib.HTTPResponse.read(self, amt) |
| 413 |
| 414 |
| 415 class AWSAuthConnection(object): |
| 416 def __init__(self, host, aws_access_key_id=None, |
| 417 aws_secret_access_key=None, |
| 418 is_secure=True, port=None, proxy=None, proxy_port=None, |
| 419 proxy_user=None, proxy_pass=None, debug=0, |
| 420 https_connection_factory=None, path='/', |
| 421 provider='aws', security_token=None, |
| 422 suppress_consec_slashes=True, |
| 423 validate_certs=True): |
| 424 """ |
| 425 :type host: str |
| 426 :param host: The host to make the connection to |
| 427 |
| 428 :keyword str aws_access_key_id: Your AWS Access Key ID (provided by |
| 429 Amazon). If none is specified, the value in your |
| 430 ``AWS_ACCESS_KEY_ID`` environmental variable is used. |
| 431 :keyword str aws_secret_access_key: Your AWS Secret Access Key |
| 432 (provided by Amazon). If none is specified, the value in your |
| 433 ``AWS_SECRET_ACCESS_KEY`` environmental variable is used. |
| 434 |
| 435 :type is_secure: boolean |
| 436 :param is_secure: Whether the connection is over SSL |
| 437 |
| 438 :type https_connection_factory: list or tuple |
| 439 :param https_connection_factory: A pair of an HTTP connection |
| 440 factory and the exceptions to catch. The factory should have |
| 441 a similar interface to L{httplib.HTTPSConnection}. |
| 442 |
| 443 :param str proxy: Address/hostname for a proxy server |
| 444 |
| 445 :type proxy_port: int |
| 446 :param proxy_port: The port to use when connecting over a proxy |
| 447 |
| 448 :type proxy_user: str |
| 449 :param proxy_user: The username to connect with on the proxy |
| 450 |
| 451 :type proxy_pass: str |
| 452 :param proxy_pass: The password to use when connection over a proxy. |
| 453 |
| 454 :type port: int |
| 455 :param port: The port to use to connect |
| 456 |
| 457 :type suppress_consec_slashes: bool |
| 458 :param suppress_consec_slashes: If provided, controls whether |
| 459 consecutive slashes will be suppressed in key paths. |
| 460 |
| 461 :type validate_certs: bool |
| 462 :param validate_certs: Controls whether SSL certificates |
| 463 will be validated or not. Defaults to True. |
| 464 """ |
| 465 self.suppress_consec_slashes = suppress_consec_slashes |
| 466 self.num_retries = 6 |
| 467 # Override passed-in is_secure setting if value was defined in config. |
| 468 if config.has_option('Boto', 'is_secure'): |
| 469 is_secure = config.getboolean('Boto', 'is_secure') |
| 470 self.is_secure = is_secure |
| 471 # Whether or not to validate server certificates. |
| 472 # The default is now to validate certificates. This can be |
| 473 # overridden in the boto config file are by passing an |
| 474 # explicit validate_certs parameter to the class constructor. |
| 475 self.https_validate_certificates = config.getbool( |
| 476 'Boto', 'https_validate_certificates', |
| 477 validate_certs) |
| 478 if self.https_validate_certificates and not HAVE_HTTPS_CONNECTION: |
| 479 raise BotoClientError( |
| 480 "SSL server certificate validation is enabled in boto " |
| 481 "configuration, but Python dependencies required to " |
| 482 "support this feature are not available. Certificate " |
| 483 "validation is only supported when running under Python " |
| 484 "2.6 or later.") |
| 485 self.ca_certificates_file = config.get_value( |
| 486 'Boto', 'ca_certificates_file', DEFAULT_CA_CERTS_FILE) |
| 487 self.handle_proxy(proxy, proxy_port, proxy_user, proxy_pass) |
| 488 # define exceptions from httplib that we want to catch and retry |
| 489 self.http_exceptions = (httplib.HTTPException, socket.error, |
| 490 socket.gaierror, httplib.BadStatusLine) |
| 491 # define subclasses of the above that are not retryable. |
| 492 self.http_unretryable_exceptions = [] |
| 493 if HAVE_HTTPS_CONNECTION: |
| 494 self.http_unretryable_exceptions.append( |
| 495 https_connection.InvalidCertificateException) |
| 496 |
| 497 # define values in socket exceptions we don't want to catch |
| 498 self.socket_exception_values = (errno.EINTR,) |
| 499 if https_connection_factory is not None: |
| 500 self.https_connection_factory = https_connection_factory[0] |
| 501 self.http_exceptions += https_connection_factory[1] |
| 502 else: |
| 503 self.https_connection_factory = None |
| 504 if (is_secure): |
| 505 self.protocol = 'https' |
| 506 else: |
| 507 self.protocol = 'http' |
| 508 self.host = host |
| 509 self.path = path |
| 510 # if the value passed in for debug |
| 511 if not isinstance(debug, (int, long)): |
| 512 debug = 0 |
| 513 self.debug = config.getint('Boto', 'debug', debug) |
| 514 if port: |
| 515 self.port = port |
| 516 else: |
| 517 self.port = PORTS_BY_SECURITY[is_secure] |
| 518 |
| 519 # Timeout used to tell httplib how long to wait for socket timeouts. |
| 520 # Default is to leave timeout unchanged, which will in turn result in |
| 521 # the socket's default global timeout being used. To specify a |
| 522 # timeout, set http_socket_timeout in Boto config. Regardless, |
| 523 # timeouts will only be applied if Python is 2.6 or greater. |
| 524 self.http_connection_kwargs = {} |
| 525 if (sys.version_info[0], sys.version_info[1]) >= (2, 6): |
| 526 if config.has_option('Boto', 'http_socket_timeout'): |
| 527 timeout = config.getint('Boto', 'http_socket_timeout') |
| 528 self.http_connection_kwargs['timeout'] = timeout |
| 529 |
| 530 if isinstance(provider, Provider): |
| 531 # Allow overriding Provider |
| 532 self.provider = provider |
| 533 else: |
| 534 self._provider_type = provider |
| 535 self.provider = Provider(self._provider_type, |
| 536 aws_access_key_id, |
| 537 aws_secret_access_key, |
| 538 security_token) |
| 539 |
| 540 # allow config file to override default host |
| 541 if self.provider.host: |
| 542 self.host = self.provider.host |
| 543 |
| 544 self._pool = ConnectionPool() |
| 545 self._connection = (self.server_name(), self.is_secure) |
| 546 self._last_rs = None |
| 547 self._auth_handler = auth.get_auth_handler( |
| 548 host, config, self.provider, self._required_auth_capability()) |
| 549 if getattr(self, 'AuthServiceName', None) is not None: |
| 550 self.auth_service_name = self.AuthServiceName |
| 551 |
| 552 def __repr__(self): |
| 553 return '%s:%s' % (self.__class__.__name__, self.host) |
| 554 |
| 555 def _required_auth_capability(self): |
| 556 return [] |
| 557 |
| 558 def _get_auth_service_name(self): |
| 559 return getattr(self._auth_handler, 'service_name') |
| 560 |
| 561 # For Sigv4, the auth_service_name/auth_region_name properties allow |
| 562 # the service_name/region_name to be explicitly set instead of being |
| 563 # derived from the endpoint url. |
| 564 def _set_auth_service_name(self, value): |
| 565 self._auth_handler.service_name = value |
| 566 auth_service_name = property(_get_auth_service_name, _set_auth_service_name) |
| 567 |
| 568 def _get_auth_region_name(self): |
| 569 return getattr(self._auth_handler, 'region_name') |
| 570 |
| 571 def _set_auth_region_name(self, value): |
| 572 self._auth_handler.region_name = value |
| 573 auth_region_name = property(_get_auth_region_name, _set_auth_region_name) |
| 574 |
| 575 def connection(self): |
| 576 return self.get_http_connection(*self._connection) |
| 577 connection = property(connection) |
| 578 |
| 579 def aws_access_key_id(self): |
| 580 return self.provider.access_key |
| 581 aws_access_key_id = property(aws_access_key_id) |
| 582 gs_access_key_id = aws_access_key_id |
| 583 access_key = aws_access_key_id |
| 584 |
| 585 def aws_secret_access_key(self): |
| 586 return self.provider.secret_key |
| 587 aws_secret_access_key = property(aws_secret_access_key) |
| 588 gs_secret_access_key = aws_secret_access_key |
| 589 secret_key = aws_secret_access_key |
| 590 |
| 591 def get_path(self, path='/'): |
| 592 # The default behavior is to suppress consecutive slashes for reasons |
| 593 # discussed at |
| 594 # https://groups.google.com/forum/#!topic/boto-dev/-ft0XPUy0y8 |
| 595 # You can override that behavior with the suppress_consec_slashes param. |
| 596 if not self.suppress_consec_slashes: |
| 597 return self.path + re.sub('^/*', "", path) |
| 598 pos = path.find('?') |
| 599 if pos >= 0: |
| 600 params = path[pos:] |
| 601 path = path[:pos] |
| 602 else: |
| 603 params = None |
| 604 if path[-1] == '/': |
| 605 need_trailing = True |
| 606 else: |
| 607 need_trailing = False |
| 608 path_elements = self.path.split('/') |
| 609 path_elements.extend(path.split('/')) |
| 610 path_elements = [p for p in path_elements if p] |
| 611 path = '/' + '/'.join(path_elements) |
| 612 if path[-1] != '/' and need_trailing: |
| 613 path += '/' |
| 614 if params: |
| 615 path = path + params |
| 616 return path |
| 617 |
| 618 def server_name(self, port=None): |
| 619 if not port: |
| 620 port = self.port |
| 621 if port == 80: |
| 622 signature_host = self.host |
| 623 else: |
| 624 # This unfortunate little hack can be attributed to |
| 625 # a difference in the 2.6 version of httplib. In old |
| 626 # versions, it would append ":443" to the hostname sent |
| 627 # in the Host header and so we needed to make sure we |
| 628 # did the same when calculating the V2 signature. In 2.6 |
| 629 # (and higher!) |
| 630 # it no longer does that. Hence, this kludge. |
| 631 if ((ON_APP_ENGINE and sys.version[:3] == '2.5') or |
| 632 sys.version[:3] in ('2.6', '2.7')) and port == 443: |
| 633 signature_host = self.host |
| 634 else: |
| 635 signature_host = '%s:%d' % (self.host, port) |
| 636 return signature_host |
| 637 |
| 638 def handle_proxy(self, proxy, proxy_port, proxy_user, proxy_pass): |
| 639 self.proxy = proxy |
| 640 self.proxy_port = proxy_port |
| 641 self.proxy_user = proxy_user |
| 642 self.proxy_pass = proxy_pass |
| 643 if 'http_proxy' in os.environ and not self.proxy: |
| 644 pattern = re.compile( |
| 645 '(?:http://)?' \ |
| 646 '(?:(?P<user>\w+):(?P<pass>.*)@)?' \ |
| 647 '(?P<host>[\w\-\.]+)' \ |
| 648 '(?::(?P<port>\d+))?' |
| 649 ) |
| 650 match = pattern.match(os.environ['http_proxy']) |
| 651 if match: |
| 652 self.proxy = match.group('host') |
| 653 self.proxy_port = match.group('port') |
| 654 self.proxy_user = match.group('user') |
| 655 self.proxy_pass = match.group('pass') |
| 656 else: |
| 657 if not self.proxy: |
| 658 self.proxy = config.get_value('Boto', 'proxy', None) |
| 659 if not self.proxy_port: |
| 660 self.proxy_port = config.get_value('Boto', 'proxy_port', None) |
| 661 if not self.proxy_user: |
| 662 self.proxy_user = config.get_value('Boto', 'proxy_user', None) |
| 663 if not self.proxy_pass: |
| 664 self.proxy_pass = config.get_value('Boto', 'proxy_pass', None) |
| 665 |
| 666 if not self.proxy_port and self.proxy: |
| 667 print "http_proxy environment variable does not specify " \ |
| 668 "a port, using default" |
| 669 self.proxy_port = self.port |
| 670 self.use_proxy = (self.proxy != None) |
| 671 |
| 672 def get_http_connection(self, host, is_secure): |
| 673 conn = self._pool.get_http_connection(host, is_secure) |
| 674 if conn is not None: |
| 675 return conn |
| 676 else: |
| 677 return self.new_http_connection(host, is_secure) |
| 678 |
| 679 def new_http_connection(self, host, is_secure): |
| 680 if self.use_proxy and not is_secure: |
| 681 host = '%s:%d' % (self.proxy, int(self.proxy_port)) |
| 682 if host is None: |
| 683 host = self.server_name() |
| 684 if is_secure: |
| 685 boto.log.debug( |
| 686 'establishing HTTPS connection: host=%s, kwargs=%s', |
| 687 host, self.http_connection_kwargs) |
| 688 if self.use_proxy: |
| 689 connection = self.proxy_ssl(host, is_secure and 443 or 80) |
| 690 elif self.https_connection_factory: |
| 691 connection = self.https_connection_factory(host) |
| 692 elif self.https_validate_certificates and HAVE_HTTPS_CONNECTION: |
| 693 connection = https_connection.CertValidatingHTTPSConnection( |
| 694 host, ca_certs=self.ca_certificates_file, |
| 695 **self.http_connection_kwargs) |
| 696 else: |
| 697 connection = httplib.HTTPSConnection(host, |
| 698 **self.http_connection_kwargs) |
| 699 else: |
| 700 boto.log.debug('establishing HTTP connection: kwargs=%s' % |
| 701 self.http_connection_kwargs) |
| 702 if self.https_connection_factory: |
| 703 # even though the factory says https, this is too handy |
| 704 # to not be able to allow overriding for http also. |
| 705 connection = self.https_connection_factory(host, |
| 706 **self.http_connection_kwargs) |
| 707 else: |
| 708 connection = httplib.HTTPConnection(host, |
| 709 **self.http_connection_kwargs) |
| 710 if self.debug > 1: |
| 711 connection.set_debuglevel(self.debug) |
| 712 # self.connection must be maintained for backwards-compatibility |
| 713 # however, it must be dynamically pulled from the connection pool |
| 714 # set a private variable which will enable that |
| 715 if host.split(':')[0] == self.host and is_secure == self.is_secure: |
| 716 self._connection = (host, is_secure) |
| 717 # Set the response class of the http connection to use our custom |
| 718 # class. |
| 719 connection.response_class = HTTPResponse |
| 720 return connection |
| 721 |
| 722 def put_http_connection(self, host, is_secure, connection): |
| 723 self._pool.put_http_connection(host, is_secure, connection) |
| 724 |
| 725 def proxy_ssl(self, host=None, port=None): |
| 726 if host and port: |
| 727 host = '%s:%d' % (host, port) |
| 728 else: |
| 729 host = '%s:%d' % (self.host, self.port) |
| 730 sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) |
| 731 try: |
| 732 sock.connect((self.proxy, int(self.proxy_port))) |
| 733 except: |
| 734 raise |
| 735 boto.log.debug("Proxy connection: CONNECT %s HTTP/1.0\r\n", host) |
| 736 sock.sendall("CONNECT %s HTTP/1.0\r\n" % host) |
| 737 sock.sendall("User-Agent: %s\r\n" % UserAgent) |
| 738 if self.proxy_user and self.proxy_pass: |
| 739 for k, v in self.get_proxy_auth_header().items(): |
| 740 sock.sendall("%s: %s\r\n" % (k, v)) |
| 741 # See discussion about this config option at |
| 742 # https://groups.google.com/forum/?fromgroups#!topic/boto-dev/teenFv
Oq2Cc |
| 743 if config.getbool('Boto', 'send_crlf_after_proxy_auth_headers', Fals
e): |
| 744 sock.sendall("\r\n") |
| 745 else: |
| 746 sock.sendall("\r\n") |
| 747 resp = httplib.HTTPResponse(sock, strict=True, debuglevel=self.debug) |
| 748 resp.begin() |
| 749 |
| 750 if resp.status != 200: |
| 751 # Fake a socket error, use a code that make it obvious it hasn't |
| 752 # been generated by the socket library |
| 753 raise socket.error(-71, |
| 754 "Error talking to HTTP proxy %s:%s: %s (%s)" % |
| 755 (self.proxy, self.proxy_port, |
| 756 resp.status, resp.reason)) |
| 757 |
| 758 # We can safely close the response, it duped the original socket |
| 759 resp.close() |
| 760 |
| 761 h = httplib.HTTPConnection(host) |
| 762 |
| 763 if self.https_validate_certificates and HAVE_HTTPS_CONNECTION: |
| 764 boto.log.debug("wrapping ssl socket for proxied connection; " |
| 765 "CA certificate file=%s", |
| 766 self.ca_certificates_file) |
| 767 key_file = self.http_connection_kwargs.get('key_file', None) |
| 768 cert_file = self.http_connection_kwargs.get('cert_file', None) |
| 769 sslSock = ssl.wrap_socket(sock, keyfile=key_file, |
| 770 certfile=cert_file, |
| 771 cert_reqs=ssl.CERT_REQUIRED, |
| 772 ca_certs=self.ca_certificates_file) |
| 773 cert = sslSock.getpeercert() |
| 774 hostname = self.host.split(':', 0)[0] |
| 775 if not https_connection.ValidateCertificateHostname(cert, hostname): |
| 776 raise https_connection.InvalidCertificateException( |
| 777 hostname, cert, 'hostname mismatch') |
| 778 else: |
| 779 # Fallback for old Python without ssl.wrap_socket |
| 780 if hasattr(httplib, 'ssl'): |
| 781 sslSock = httplib.ssl.SSLSocket(sock) |
| 782 else: |
| 783 sslSock = socket.ssl(sock, None, None) |
| 784 sslSock = httplib.FakeSocket(sock, sslSock) |
| 785 |
| 786 # This is a bit unclean |
| 787 h.sock = sslSock |
| 788 return h |
| 789 |
| 790 def prefix_proxy_to_path(self, path, host=None): |
| 791 path = self.protocol + '://' + (host or self.server_name()) + path |
| 792 return path |
| 793 |
| 794 def get_proxy_auth_header(self): |
| 795 auth = base64.encodestring(self.proxy_user + ':' + self.proxy_pass) |
| 796 return {'Proxy-Authorization': 'Basic %s' % auth} |
| 797 |
| 798 def _mexe(self, request, sender=None, override_num_retries=None, |
| 799 retry_handler=None): |
| 800 """ |
| 801 mexe - Multi-execute inside a loop, retrying multiple times to handle |
| 802 transient Internet errors by simply trying again. |
| 803 Also handles redirects. |
| 804 |
| 805 This code was inspired by the S3Utils classes posted to the boto-users |
| 806 Google group by Larry Bates. Thanks! |
| 807 |
| 808 """ |
| 809 boto.log.debug('Method: %s' % request.method) |
| 810 boto.log.debug('Path: %s' % request.path) |
| 811 boto.log.debug('Data: %s' % request.body) |
| 812 boto.log.debug('Headers: %s' % request.headers) |
| 813 boto.log.debug('Host: %s' % request.host) |
| 814 response = None |
| 815 body = None |
| 816 e = None |
| 817 if override_num_retries is None: |
| 818 num_retries = config.getint('Boto', 'num_retries', self.num_retries) |
| 819 else: |
| 820 num_retries = override_num_retries |
| 821 i = 0 |
| 822 connection = self.get_http_connection(request.host, self.is_secure) |
| 823 while i <= num_retries: |
| 824 # Use binary exponential backoff to desynchronize client requests. |
| 825 next_sleep = random.random() * (2 ** i) |
| 826 try: |
| 827 # we now re-sign each request before it is retried |
| 828 boto.log.debug('Token: %s' % self.provider.security_token) |
| 829 request.authorize(connection=self) |
| 830 if callable(sender): |
| 831 response = sender(connection, request.method, request.path, |
| 832 request.body, request.headers) |
| 833 else: |
| 834 connection.request(request.method, request.path, |
| 835 request.body, request.headers) |
| 836 response = connection.getresponse() |
| 837 location = response.getheader('location') |
| 838 # -- gross hack -- |
| 839 # httplib gets confused with chunked responses to HEAD requests |
| 840 # so I have to fake it out |
| 841 if request.method == 'HEAD' and getattr(response, |
| 842 'chunked', False): |
| 843 response.chunked = 0 |
| 844 if callable(retry_handler): |
| 845 status = retry_handler(response, i, next_sleep) |
| 846 if status: |
| 847 msg, i, next_sleep = status |
| 848 if msg: |
| 849 boto.log.debug(msg) |
| 850 time.sleep(next_sleep) |
| 851 continue |
| 852 if response.status == 500 or response.status == 503: |
| 853 msg = 'Received %d response. ' % response.status |
| 854 msg += 'Retrying in %3.1f seconds' % next_sleep |
| 855 boto.log.debug(msg) |
| 856 body = response.read() |
| 857 elif response.status < 300 or response.status >= 400 or \ |
| 858 not location: |
| 859 self.put_http_connection(request.host, self.is_secure, |
| 860 connection) |
| 861 return response |
| 862 else: |
| 863 scheme, request.host, request.path, \ |
| 864 params, query, fragment = urlparse.urlparse(location) |
| 865 if query: |
| 866 request.path += '?' + query |
| 867 msg = 'Redirecting: %s' % scheme + '://' |
| 868 msg += request.host + request.path |
| 869 boto.log.debug(msg) |
| 870 connection = self.get_http_connection(request.host, |
| 871 scheme == 'https') |
| 872 response = None |
| 873 continue |
| 874 except self.http_exceptions, e: |
| 875 for unretryable in self.http_unretryable_exceptions: |
| 876 if isinstance(e, unretryable): |
| 877 boto.log.debug( |
| 878 'encountered unretryable %s exception, re-raising' % |
| 879 e.__class__.__name__) |
| 880 raise e |
| 881 boto.log.debug('encountered %s exception, reconnecting' % \ |
| 882 e.__class__.__name__) |
| 883 connection = self.new_http_connection(request.host, |
| 884 self.is_secure) |
| 885 time.sleep(next_sleep) |
| 886 i += 1 |
| 887 # If we made it here, it's because we have exhausted our retries |
| 888 # and stil haven't succeeded. So, if we have a response object, |
| 889 # use it to raise an exception. |
| 890 # Otherwise, raise the exception that must have already h#appened. |
| 891 if response: |
| 892 raise BotoServerError(response.status, response.reason, body) |
| 893 elif e: |
| 894 raise e |
| 895 else: |
| 896 msg = 'Please report this exception as a Boto Issue!' |
| 897 raise BotoClientError(msg) |
| 898 |
| 899 def build_base_http_request(self, method, path, auth_path, |
| 900 params=None, headers=None, data='', host=None): |
| 901 path = self.get_path(path) |
| 902 if auth_path is not None: |
| 903 auth_path = self.get_path(auth_path) |
| 904 if params == None: |
| 905 params = {} |
| 906 else: |
| 907 params = params.copy() |
| 908 if headers == None: |
| 909 headers = {} |
| 910 else: |
| 911 headers = headers.copy() |
| 912 host = host or self.host |
| 913 if self.use_proxy: |
| 914 if not auth_path: |
| 915 auth_path = path |
| 916 path = self.prefix_proxy_to_path(path, host) |
| 917 if self.proxy_user and self.proxy_pass and not self.is_secure: |
| 918 # If is_secure, we don't have to set the proxy authentication |
| 919 # header here, we did that in the CONNECT to the proxy. |
| 920 headers.update(self.get_proxy_auth_header()) |
| 921 return HTTPRequest(method, self.protocol, host, self.port, |
| 922 path, auth_path, params, headers, data) |
| 923 |
| 924 def make_request(self, method, path, headers=None, data='', host=None, |
| 925 auth_path=None, sender=None, override_num_retries=None, |
| 926 params=None): |
| 927 """Makes a request to the server, with stock multiple-retry logic.""" |
| 928 if params is None: |
| 929 params = {} |
| 930 http_request = self.build_base_http_request(method, path, auth_path, |
| 931 params, headers, data, host) |
| 932 return self._mexe(http_request, sender, override_num_retries) |
| 933 |
| 934 def close(self): |
| 935 """(Optional) Close any open HTTP connections. This is non-destructive, |
| 936 and making a new request will open a connection again.""" |
| 937 |
| 938 boto.log.debug('closing all HTTP connections') |
| 939 self._connection = None # compat field |
| 940 |
| 941 |
| 942 class AWSQueryConnection(AWSAuthConnection): |
| 943 |
| 944 APIVersion = '' |
| 945 ResponseError = BotoServerError |
| 946 |
| 947 def __init__(self, aws_access_key_id=None, aws_secret_access_key=None, |
| 948 is_secure=True, port=None, proxy=None, proxy_port=None, |
| 949 proxy_user=None, proxy_pass=None, host=None, debug=0, |
| 950 https_connection_factory=None, path='/', security_token=None, |
| 951 validate_certs=True): |
| 952 AWSAuthConnection.__init__(self, host, aws_access_key_id, |
| 953 aws_secret_access_key, |
| 954 is_secure, port, proxy, |
| 955 proxy_port, proxy_user, proxy_pass, |
| 956 debug, https_connection_factory, path, |
| 957 security_token=security_token, |
| 958 validate_certs=validate_certs) |
| 959 |
| 960 def _required_auth_capability(self): |
| 961 return [] |
| 962 |
| 963 def get_utf8_value(self, value): |
| 964 return boto.utils.get_utf8_value(value) |
| 965 |
| 966 def make_request(self, action, params=None, path='/', verb='GET'): |
| 967 http_request = self.build_base_http_request(verb, path, None, |
| 968 params, {}, '', |
| 969 self.server_name()) |
| 970 if action: |
| 971 http_request.params['Action'] = action |
| 972 if self.APIVersion: |
| 973 http_request.params['Version'] = self.APIVersion |
| 974 return self._mexe(http_request) |
| 975 |
| 976 def build_list_params(self, params, items, label): |
| 977 if isinstance(items, basestring): |
| 978 items = [items] |
| 979 for i in range(1, len(items) + 1): |
| 980 params['%s.%d' % (label, i)] = items[i - 1] |
| 981 |
| 982 def build_complex_list_params(self, params, items, label, names): |
| 983 """Serialize a list of structures. |
| 984 |
| 985 For example:: |
| 986 |
| 987 items = [('foo', 'bar', 'baz'), ('foo2', 'bar2', 'baz2')] |
| 988 label = 'ParamName.member' |
| 989 names = ('One', 'Two', 'Three') |
| 990 self.build_complex_list_params(params, items, label, names) |
| 991 |
| 992 would result in the params dict being updated with these params:: |
| 993 |
| 994 ParamName.member.1.One = foo |
| 995 ParamName.member.1.Two = bar |
| 996 ParamName.member.1.Three = baz |
| 997 |
| 998 ParamName.member.2.One = foo2 |
| 999 ParamName.member.2.Two = bar2 |
| 1000 ParamName.member.2.Three = baz2 |
| 1001 |
| 1002 :type params: dict |
| 1003 :param params: The params dict. The complex list params |
| 1004 will be added to this dict. |
| 1005 |
| 1006 :type items: list of tuples |
| 1007 :param items: The list to serialize. |
| 1008 |
| 1009 :type label: string |
| 1010 :param label: The prefix to apply to the parameter. |
| 1011 |
| 1012 :type names: tuple of strings |
| 1013 :param names: The names associated with each tuple element. |
| 1014 |
| 1015 """ |
| 1016 for i, item in enumerate(items, 1): |
| 1017 current_prefix = '%s.%s' % (label, i) |
| 1018 for key, value in zip(names, item): |
| 1019 full_key = '%s.%s' % (current_prefix, key) |
| 1020 params[full_key] = value |
| 1021 |
| 1022 # generics |
| 1023 |
| 1024 def get_list(self, action, params, markers, path='/', |
| 1025 parent=None, verb='GET'): |
| 1026 if not parent: |
| 1027 parent = self |
| 1028 response = self.make_request(action, params, path, verb) |
| 1029 body = response.read() |
| 1030 boto.log.debug(body) |
| 1031 if not body: |
| 1032 boto.log.error('Null body %s' % body) |
| 1033 raise self.ResponseError(response.status, response.reason, body) |
| 1034 elif response.status == 200: |
| 1035 rs = ResultSet(markers) |
| 1036 h = boto.handler.XmlHandler(rs, parent) |
| 1037 xml.sax.parseString(body, h) |
| 1038 return rs |
| 1039 else: |
| 1040 boto.log.error('%s %s' % (response.status, response.reason)) |
| 1041 boto.log.error('%s' % body) |
| 1042 raise self.ResponseError(response.status, response.reason, body) |
| 1043 |
| 1044 def get_object(self, action, params, cls, path='/', |
| 1045 parent=None, verb='GET'): |
| 1046 if not parent: |
| 1047 parent = self |
| 1048 response = self.make_request(action, params, path, verb) |
| 1049 body = response.read() |
| 1050 boto.log.debug(body) |
| 1051 if not body: |
| 1052 boto.log.error('Null body %s' % body) |
| 1053 raise self.ResponseError(response.status, response.reason, body) |
| 1054 elif response.status == 200: |
| 1055 obj = cls(parent) |
| 1056 h = boto.handler.XmlHandler(obj, parent) |
| 1057 xml.sax.parseString(body, h) |
| 1058 return obj |
| 1059 else: |
| 1060 boto.log.error('%s %s' % (response.status, response.reason)) |
| 1061 boto.log.error('%s' % body) |
| 1062 raise self.ResponseError(response.status, response.reason, body) |
| 1063 |
| 1064 def get_status(self, action, params, path='/', parent=None, verb='GET'): |
| 1065 if not parent: |
| 1066 parent = self |
| 1067 response = self.make_request(action, params, path, verb) |
| 1068 body = response.read() |
| 1069 boto.log.debug(body) |
| 1070 if not body: |
| 1071 boto.log.error('Null body %s' % body) |
| 1072 raise self.ResponseError(response.status, response.reason, body) |
| 1073 elif response.status == 200: |
| 1074 rs = ResultSet() |
| 1075 h = boto.handler.XmlHandler(rs, parent) |
| 1076 xml.sax.parseString(body, h) |
| 1077 return rs.status |
| 1078 else: |
| 1079 boto.log.error('%s %s' % (response.status, response.reason)) |
| 1080 boto.log.error('%s' % body) |
| 1081 raise self.ResponseError(response.status, response.reason, body) |
OLD | NEW |