12 import arvados_testutil as tutil
13 import run_test_server
15 class KeepTestCase(run_test_server.TestCaseWithServers):
21 super(KeepTestCase, cls).setUpClass()
22 run_test_server.authorize_with("admin")
23 cls.api_client = arvados.api('v1')
24 cls.keep_client = arvados.KeepClient(api_client=cls.api_client,
25 proxy='', local_store='')
27 def test_KeepBasicRWTest(self):
28 foo_locator = self.keep_client.put('foo')
29 self.assertRegexpMatches(
31 '^acbd18db4cc2f85cedef654fccc4a4d8\+3',
32 'wrong md5 hash from Keep.put("foo"): ' + foo_locator)
33 self.assertEqual(self.keep_client.get(foo_locator),
35 'wrong content from Keep.get(md5("foo"))')
37 def test_KeepBinaryRWTest(self):
38 blob_str = '\xff\xfe\xf7\x00\x01\x02'
39 blob_locator = self.keep_client.put(blob_str)
40 self.assertRegexpMatches(
42 '^7fc7c53b45e53926ba52821140fef396\+6',
43 ('wrong locator from Keep.put(<binarydata>):' + blob_locator))
44 self.assertEqual(self.keep_client.get(blob_locator),
46 'wrong content from Keep.get(md5(<binarydata>))')
48 def test_KeepLongBinaryRWTest(self):
49 blob_str = '\xff\xfe\xfd\xfc\x00\x01\x02\x03'
51 blob_str = blob_str + blob_str
52 blob_locator = self.keep_client.put(blob_str)
53 self.assertRegexpMatches(
55 '^84d90fc0d8175dd5dcfab04b999bc956\+67108864',
56 ('wrong locator from Keep.put(<binarydata>): ' + blob_locator))
57 self.assertEqual(self.keep_client.get(blob_locator),
59 'wrong content from Keep.get(md5(<binarydata>))')
61 def test_KeepSingleCopyRWTest(self):
62 blob_str = '\xff\xfe\xfd\xfc\x00\x01\x02\x03'
63 blob_locator = self.keep_client.put(blob_str, copies=1)
64 self.assertRegexpMatches(
66 '^c902006bc98a3eb4a3663b65ab4a6fab\+8',
67 ('wrong locator from Keep.put(<binarydata>): ' + blob_locator))
68 self.assertEqual(self.keep_client.get(blob_locator),
70 'wrong content from Keep.get(md5(<binarydata>))')
72 def test_KeepEmptyCollectionTest(self):
73 blob_locator = self.keep_client.put('', copies=1)
74 self.assertRegexpMatches(
76 '^d41d8cd98f00b204e9800998ecf8427e\+0',
77 ('wrong locator from Keep.put(""): ' + blob_locator))
80 class KeepPermissionTestCase(run_test_server.TestCaseWithServers):
82 KEEP_SERVER = {'blob_signing_key': 'abcdefghijk0123456789',
83 'enforce_permissions': True}
85 def test_KeepBasicRWTest(self):
86 run_test_server.authorize_with('active')
87 keep_client = arvados.KeepClient()
88 foo_locator = keep_client.put('foo')
89 self.assertRegexpMatches(
91 r'^acbd18db4cc2f85cedef654fccc4a4d8\+3\+A[a-f0-9]+@[a-f0-9]+$',
92 'invalid locator from Keep.put("foo"): ' + foo_locator)
93 self.assertEqual(keep_client.get(foo_locator),
95 'wrong content from Keep.get(md5("foo"))')
97 # GET with an unsigned locator => NotFound
98 bar_locator = keep_client.put('bar')
99 unsigned_bar_locator = "37b51d194a7513e45b56f6524f2d51f2+3"
100 self.assertRegexpMatches(
102 r'^37b51d194a7513e45b56f6524f2d51f2\+3\+A[a-f0-9]+@[a-f0-9]+$',
103 'invalid locator from Keep.put("bar"): ' + bar_locator)
104 self.assertRaises(arvados.errors.NotFoundError,
106 unsigned_bar_locator)
108 # GET from a different user => NotFound
109 run_test_server.authorize_with('spectator')
110 self.assertRaises(arvados.errors.NotFoundError,
114 # Unauthenticated GET for a signed locator => NotFound
115 # Unauthenticated GET for an unsigned locator => NotFound
116 keep_client.api_token = ''
117 self.assertRaises(arvados.errors.NotFoundError,
120 self.assertRaises(arvados.errors.NotFoundError,
122 unsigned_bar_locator)
125 # KeepOptionalPermission: starts Keep with --permission-key-file
126 # but not --enforce-permissions (i.e. generate signatures on PUT
127 # requests, but do not require them for GET requests)
129 # All of these requests should succeed when permissions are optional:
130 # * authenticated request, signed locator
131 # * authenticated request, unsigned locator
132 # * unauthenticated request, signed locator
133 # * unauthenticated request, unsigned locator
134 class KeepOptionalPermission(run_test_server.TestCaseWithServers):
136 KEEP_SERVER = {'blob_signing_key': 'abcdefghijk0123456789',
137 'enforce_permissions': False}
141 super(KeepOptionalPermission, cls).setUpClass()
142 run_test_server.authorize_with("admin")
143 cls.api_client = arvados.api('v1')
146 super(KeepOptionalPermission, self).setUp()
147 self.keep_client = arvados.KeepClient(api_client=self.api_client,
148 proxy='', local_store='')
150 def _put_foo_and_check(self):
151 signed_locator = self.keep_client.put('foo')
152 self.assertRegexpMatches(
154 r'^acbd18db4cc2f85cedef654fccc4a4d8\+3\+A[a-f0-9]+@[a-f0-9]+$',
155 'invalid locator from Keep.put("foo"): ' + signed_locator)
156 return signed_locator
158 def test_KeepAuthenticatedSignedTest(self):
159 signed_locator = self._put_foo_and_check()
160 self.assertEqual(self.keep_client.get(signed_locator),
162 'wrong content from Keep.get(md5("foo"))')
164 def test_KeepAuthenticatedUnsignedTest(self):
165 signed_locator = self._put_foo_and_check()
166 self.assertEqual(self.keep_client.get("acbd18db4cc2f85cedef654fccc4a4d8"),
168 'wrong content from Keep.get(md5("foo"))')
170 def test_KeepUnauthenticatedSignedTest(self):
171 # Check that signed GET requests work even when permissions
172 # enforcement is off.
173 signed_locator = self._put_foo_and_check()
174 self.keep_client.api_token = ''
175 self.assertEqual(self.keep_client.get(signed_locator),
177 'wrong content from Keep.get(md5("foo"))')
179 def test_KeepUnauthenticatedUnsignedTest(self):
180 # Since --enforce-permissions is not in effect, GET requests
181 # need not be authenticated.
182 signed_locator = self._put_foo_and_check()
183 self.keep_client.api_token = ''
184 self.assertEqual(self.keep_client.get("acbd18db4cc2f85cedef654fccc4a4d8"),
186 'wrong content from Keep.get(md5("foo"))')
189 class KeepProxyTestCase(run_test_server.TestCaseWithServers):
192 KEEP_PROXY_SERVER = {}
196 super(KeepProxyTestCase, cls).setUpClass()
197 run_test_server.authorize_with('active')
198 cls.api_client = arvados.api('v1')
201 arvados.config.settings().pop('ARVADOS_EXTERNAL_CLIENT', None)
202 super(KeepProxyTestCase, self).tearDown()
204 def test_KeepProxyTest1(self):
205 # Will use ARVADOS_KEEP_PROXY environment variable that is set by
207 keep_client = arvados.KeepClient(api_client=self.api_client,
209 baz_locator = keep_client.put('baz')
210 self.assertRegexpMatches(
212 '^73feffa4b7f6bb68e44cf984c85f6e88\+3',
213 'wrong md5 hash from Keep.put("baz"): ' + baz_locator)
214 self.assertEqual(keep_client.get(baz_locator),
216 'wrong content from Keep.get(md5("baz"))')
217 self.assertTrue(keep_client.using_proxy)
219 def test_KeepProxyTest2(self):
220 # Don't instantiate the proxy directly, but set the X-External-Client
221 # header. The API server should direct us to the proxy.
222 arvados.config.settings()['ARVADOS_EXTERNAL_CLIENT'] = 'true'
223 keep_client = arvados.KeepClient(api_client=self.api_client,
224 proxy='', local_store='')
225 baz_locator = keep_client.put('baz2')
226 self.assertRegexpMatches(
228 '^91f372a266fe2bf2823cb8ec7fda31ce\+4',
229 'wrong md5 hash from Keep.put("baz2"): ' + baz_locator)
230 self.assertEqual(keep_client.get(baz_locator),
232 'wrong content from Keep.get(md5("baz2"))')
233 self.assertTrue(keep_client.using_proxy)
236 class KeepClientServiceTestCase(unittest.TestCase, tutil.ApiClientMock):
237 def get_service_roots(self, api_client):
238 keep_client = arvados.KeepClient(api_client=api_client)
239 services = keep_client.weighted_service_roots('000000')
240 return [urlparse.urlparse(url) for url in sorted(services)]
242 def test_ssl_flag_respected_in_roots(self):
243 for ssl_flag in [False, True]:
244 services = self.get_service_roots(self.mock_keep_services(
245 service_ssl_flag=ssl_flag))
247 ('https' if ssl_flag else 'http'), services[0].scheme)
249 def test_correct_ports_with_ipv6_addresses(self):
250 service = self.get_service_roots(self.mock_keep_services(
251 service_type='proxy', service_host='100::1', service_port=10, count=1))[0]
252 self.assertEqual('100::1', service.hostname)
253 self.assertEqual(10, service.port)
255 # test_get_timeout and test_put_timeout test that
256 # KeepClient.get and KeepClient.put use the appropriate timeouts
257 # when connected directly to a Keep server (i.e. non-proxy timeout)
259 def test_get_timeout(self):
260 api_client = self.mock_keep_services(count=1)
261 keep_client = arvados.KeepClient(api_client=api_client)
262 force_timeout = [socket.timeout("timed out")]
263 with mock.patch('requests.get', side_effect=force_timeout) as mock_request:
264 with self.assertRaises(arvados.errors.KeepReadError):
265 keep_client.get('ffffffffffffffffffffffffffffffff')
266 self.assertTrue(mock_request.called)
268 arvados.KeepClient.DEFAULT_TIMEOUT,
269 mock_request.call_args[1]['timeout'])
271 def test_put_timeout(self):
272 api_client = self.mock_keep_services(count=1)
273 keep_client = arvados.KeepClient(api_client=api_client)
274 force_timeout = [socket.timeout("timed out")]
275 with mock.patch('requests.put', side_effect=force_timeout) as mock_request:
276 with self.assertRaises(arvados.errors.KeepWriteError):
277 keep_client.put('foo')
278 self.assertTrue(mock_request.called)
280 arvados.KeepClient.DEFAULT_TIMEOUT,
281 mock_request.call_args[1]['timeout'])
283 def test_proxy_get_timeout(self):
284 # Force a timeout, verifying that the requests.get or
285 # requests.put method was called with the proxy_timeout
286 # setting rather than the default timeout.
287 api_client = self.mock_keep_services(service_type='proxy', count=1)
288 keep_client = arvados.KeepClient(api_client=api_client)
289 force_timeout = [socket.timeout("timed out")]
290 with mock.patch('requests.get', side_effect=force_timeout) as mock_request:
291 with self.assertRaises(arvados.errors.KeepReadError):
292 keep_client.get('ffffffffffffffffffffffffffffffff')
293 self.assertTrue(mock_request.called)
295 arvados.KeepClient.DEFAULT_PROXY_TIMEOUT,
296 mock_request.call_args[1]['timeout'])
298 def test_proxy_put_timeout(self):
299 # Force a timeout, verifying that the requests.get or
300 # requests.put method was called with the proxy_timeout
301 # setting rather than the default timeout.
302 api_client = self.mock_keep_services(service_type='proxy', count=1)
303 keep_client = arvados.KeepClient(api_client=api_client)
304 force_timeout = [socket.timeout("timed out")]
305 with mock.patch('requests.put', side_effect=force_timeout) as mock_request:
306 with self.assertRaises(arvados.errors.KeepWriteError):
307 keep_client.put('foo')
308 self.assertTrue(mock_request.called)
310 arvados.KeepClient.DEFAULT_PROXY_TIMEOUT,
311 mock_request.call_args[1]['timeout'])
313 def test_probe_order_reference_set(self):
314 # expected_order[i] is the probe order for
315 # hash=md5(sprintf("%064x",i)) where there are 16 services
316 # with uuid sprintf("anything-%015x",j) with j in 0..15. E.g.,
317 # the first probe for the block consisting of 64 "0"
318 # characters is the service whose uuid is
319 # "zzzzz-bi6l4-000000000000003", so expected_order[0][0]=='3'.
321 list('3eab2d5fc9681074'),
322 list('097dba52e648f1c3'),
323 list('c5b4e023f8a7d691'),
324 list('9d81c02e76a3bf54'),
327 hashlib.md5("{:064x}".format(x)).hexdigest()
328 for x in range(len(expected_order))]
329 api_client = self.mock_keep_services(count=16)
330 keep_client = arvados.KeepClient(api_client=api_client)
331 for i, hash in enumerate(hashes):
332 roots = keep_client.weighted_service_roots(hash)
334 re.search(r'//\[?keep0x([0-9a-f]+)', root).group(1)
336 self.assertEqual(expected_order[i], got_order)
338 def test_probe_waste_adding_one_server(self):
340 hashlib.md5("{:064x}".format(x)).hexdigest() for x in range(100)]
341 initial_services = 12
342 api_client = self.mock_keep_services(count=initial_services)
343 keep_client = arvados.KeepClient(api_client=api_client)
345 keep_client.weighted_service_roots(hash) for hash in hashes]
346 for added_services in range(1, 12):
347 api_client = self.mock_keep_services(count=initial_services+added_services)
348 keep_client = arvados.KeepClient(api_client=api_client)
350 for hash_index in range(len(hashes)):
351 probe_after = keep_client.weighted_service_roots(
353 penalty = probe_after.index(probes_before[hash_index][0])
354 self.assertLessEqual(penalty, added_services)
355 total_penalty += penalty
356 # Average penalty per block should not exceed
357 # N(added)/N(orig) by more than 20%, and should get closer
358 # to the ideal as we add data points.
361 len(hashes) / initial_services)
364 (120 - added_services)/100)
366 expect_penalty * 8/10)
368 min_penalty <= total_penalty <= max_penalty,
369 "With {}+{} services, {} blocks, penalty {} but expected {}..{}".format(
377 def check_64_zeros_error_order(self, verb, exc_class):
380 data = hashlib.md5(data).hexdigest() + '+1234'
381 # Arbitrary port number:
382 aport = random.randint(1024,65535)
383 api_client = self.mock_keep_services(service_port=aport, count=16)
384 keep_client = arvados.KeepClient(api_client=api_client)
385 with mock.patch('requests.' + verb,
386 side_effect=socket.timeout) as req_mock, \
387 self.assertRaises(exc_class) as err_check:
388 getattr(keep_client, verb)(data)
389 urls = [urlparse.urlparse(url)
390 for url in err_check.exception.service_errors()]
391 self.assertEqual([('keep0x' + c, aport) for c in '3eab2d5fc9681074'],
392 [(url.hostname, url.port) for url in urls])
394 def test_get_error_shows_probe_order(self):
395 self.check_64_zeros_error_order('get', arvados.errors.KeepReadError)
397 def test_put_error_shows_probe_order(self):
398 self.check_64_zeros_error_order('put', arvados.errors.KeepWriteError)
400 def check_no_services_error(self, verb, exc_class):
401 api_client = mock.MagicMock(name='api_client')
402 api_client.keep_services().accessible().execute.side_effect = (
403 arvados.errors.ApiError)
404 keep_client = arvados.KeepClient(api_client=api_client)
405 with self.assertRaises(exc_class) as err_check:
406 getattr(keep_client, verb)('d41d8cd98f00b204e9800998ecf8427e+0')
407 self.assertEqual(0, len(err_check.exception.service_errors()))
409 def test_get_error_with_no_services(self):
410 self.check_no_services_error('get', arvados.errors.KeepReadError)
412 def test_put_error_with_no_services(self):
413 self.check_no_services_error('put', arvados.errors.KeepWriteError)
415 def check_errors_from_last_retry(self, verb, exc_class):
416 api_client = self.mock_keep_services(count=2)
417 keep_client = arvados.KeepClient(api_client=api_client)
418 req_mock = getattr(tutil, 'mock_{}_responses'.format(verb))(
419 "retry error reporting test", 500, 500, 403, 403)
420 with req_mock, tutil.skip_sleep, \
421 self.assertRaises(exc_class) as err_check:
422 getattr(keep_client, verb)('d41d8cd98f00b204e9800998ecf8427e+0',
424 self.assertEqual([403, 403], [
425 getattr(error, 'status_code', None)
426 for error in err_check.exception.service_errors().itervalues()])
428 def test_get_error_reflects_last_retry(self):
429 self.check_errors_from_last_retry('get', arvados.errors.KeepReadError)
431 def test_put_error_reflects_last_retry(self):
432 self.check_errors_from_last_retry('put', arvados.errors.KeepWriteError)
434 def test_put_error_does_not_include_successful_puts(self):
435 data = 'partial failure test'
436 data_loc = '{}+{}'.format(hashlib.md5(data).hexdigest(), len(data))
437 api_client = self.mock_keep_services(count=3)
438 keep_client = arvados.KeepClient(api_client=api_client)
439 with tutil.mock_put_responses(data_loc, 200, 500, 500) as req_mock, \
440 self.assertRaises(arvados.errors.KeepWriteError) as exc_check:
441 keep_client.put(data)
442 self.assertEqual(2, len(exc_check.exception.service_errors()))
445 class KeepClientRetryTestMixin(object):
446 # Testing with a local Keep store won't exercise the retry behavior.
447 # Instead, our strategy is:
448 # * Create a client with one proxy specified (pointed at a black
449 # hole), so there's no need to instantiate an API client, and
450 # all HTTP requests come from one place.
451 # * Mock httplib's request method to provide simulated responses.
452 # This lets us test the retry logic extensively without relying on any
453 # supporting servers, and prevents side effects in case something hiccups.
454 # To use this mixin, define DEFAULT_EXPECT, DEFAULT_EXCEPTION, and
457 # Test classes must define TEST_PATCHER to a method that mocks
458 # out appropriate methods in the client.
460 PROXY_ADDR = 'http://[%s]:65535/' % (tutil.TEST_HOST,)
461 TEST_DATA = 'testdata'
462 TEST_LOCATOR = 'ef654c40ab4f1747fc699915d4f70902+8'
465 self.client_kwargs = {'proxy': self.PROXY_ADDR, 'local_store': ''}
467 def new_client(self, **caller_kwargs):
468 kwargs = self.client_kwargs.copy()
469 kwargs.update(caller_kwargs)
470 return arvados.KeepClient(**kwargs)
472 def run_method(self, *args, **kwargs):
473 raise NotImplementedError("test subclasses must define run_method")
475 def check_success(self, expected=None, *args, **kwargs):
477 expected = self.DEFAULT_EXPECT
478 self.assertEqual(expected, self.run_method(*args, **kwargs))
480 def check_exception(self, error_class=None, *args, **kwargs):
481 if error_class is None:
482 error_class = self.DEFAULT_EXCEPTION
483 self.assertRaises(error_class, self.run_method, *args, **kwargs)
485 def test_immediate_success(self):
486 with self.TEST_PATCHER(self.DEFAULT_EXPECT, 200):
489 def test_retry_then_success(self):
490 with self.TEST_PATCHER(self.DEFAULT_EXPECT, 500, 200):
491 self.check_success(num_retries=3)
493 def test_no_default_retry(self):
494 with self.TEST_PATCHER(self.DEFAULT_EXPECT, 500, 200):
495 self.check_exception()
497 def test_no_retry_after_permanent_error(self):
498 with self.TEST_PATCHER(self.DEFAULT_EXPECT, 403, 200):
499 self.check_exception(num_retries=3)
501 def test_error_after_retries_exhausted(self):
502 with self.TEST_PATCHER(self.DEFAULT_EXPECT, 500, 500, 200):
503 self.check_exception(num_retries=1)
505 def test_num_retries_instance_fallback(self):
506 self.client_kwargs['num_retries'] = 3
507 with self.TEST_PATCHER(self.DEFAULT_EXPECT, 500, 200):
512 class KeepClientRetryGetTestCase(KeepClientRetryTestMixin, unittest.TestCase):
513 DEFAULT_EXPECT = KeepClientRetryTestMixin.TEST_DATA
514 DEFAULT_EXCEPTION = arvados.errors.KeepReadError
515 HINTED_LOCATOR = KeepClientRetryTestMixin.TEST_LOCATOR + '+K@xyzzy'
516 TEST_PATCHER = staticmethod(tutil.mock_get_responses)
518 def run_method(self, locator=KeepClientRetryTestMixin.TEST_LOCATOR,
520 return self.new_client().get(locator, *args, **kwargs)
522 def test_specific_exception_when_not_found(self):
523 with tutil.mock_get_responses(self.DEFAULT_EXPECT, 404, 200):
524 self.check_exception(arvados.errors.NotFoundError, num_retries=3)
526 def test_general_exception_with_mixed_errors(self):
527 # get should raise a NotFoundError if no server returns the block,
528 # and a high threshold of servers report that it's not found.
529 # This test rigs up 50/50 disagreement between two servers, and
530 # checks that it does not become a NotFoundError.
531 client = self.new_client()
532 with tutil.mock_get_responses(self.DEFAULT_EXPECT, 404, 500):
533 with self.assertRaises(arvados.errors.KeepReadError) as exc_check:
534 client.get(self.HINTED_LOCATOR)
535 self.assertNotIsInstance(
536 exc_check.exception, arvados.errors.NotFoundError,
537 "mixed errors raised NotFoundError")
539 def test_hint_server_can_succeed_without_retries(self):
540 with tutil.mock_get_responses(self.DEFAULT_EXPECT, 404, 200, 500):
541 self.check_success(locator=self.HINTED_LOCATOR)
543 def test_try_next_server_after_timeout(self):
545 socket.timeout("timed out"),
546 tutil.fake_requests_response(200, self.DEFAULT_EXPECT)]
547 with mock.patch('requests.get',
548 side_effect=iter(side_effects)):
549 self.check_success(locator=self.HINTED_LOCATOR)
551 def test_retry_data_with_wrong_checksum(self):
552 side_effects = (tutil.fake_requests_response(200, s)
553 for s in ['baddata', self.TEST_DATA])
554 with mock.patch('requests.get', side_effect=side_effects):
555 self.check_success(locator=self.HINTED_LOCATOR)
559 class KeepClientRetryPutTestCase(KeepClientRetryTestMixin, unittest.TestCase):
560 DEFAULT_EXPECT = KeepClientRetryTestMixin.TEST_LOCATOR
561 DEFAULT_EXCEPTION = arvados.errors.KeepWriteError
562 TEST_PATCHER = staticmethod(tutil.mock_put_responses)
564 def run_method(self, data=KeepClientRetryTestMixin.TEST_DATA,
565 copies=1, *args, **kwargs):
566 return self.new_client().put(data, copies, *args, **kwargs)
568 def test_do_not_send_multiple_copies_to_same_server(self):
569 with tutil.mock_put_responses(self.DEFAULT_EXPECT, 200):
570 self.check_exception(copies=2, num_retries=3)