X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/afdab87268e4ee2f19f92e02b219696081959378..5b1602a521d3f6689f5988cdc2b25b74ff9fc1e1:/sdk/python/tests/test_keep_client.py diff --git a/sdk/python/tests/test_keep_client.py b/sdk/python/tests/test_keep_client.py index 982e4b44a8..be13c55048 100644 --- a/sdk/python/tests/test_keep_client.py +++ b/sdk/python/tests/test_keep_client.py @@ -1,6 +1,7 @@ import hashlib import mock import os +import random import re import socket import unittest @@ -75,6 +76,21 @@ class KeepTestCase(run_test_server.TestCaseWithServers): '^d41d8cd98f00b204e9800998ecf8427e\+0', ('wrong locator from Keep.put(""): ' + blob_locator)) + def test_unicode_must_be_ascii(self): + # If unicode type, must only consist of valid ASCII + foo_locator = self.keep_client.put(u'foo') + self.assertRegexpMatches( + foo_locator, + '^acbd18db4cc2f85cedef654fccc4a4d8\+3', + 'wrong md5 hash from Keep.put("foo"): ' + foo_locator) + + with self.assertRaises(UnicodeEncodeError): + # Error if it is not ASCII + self.keep_client.put(u'\xe2') + + with self.assertRaises(arvados.errors.ArgumentError): + # Must be a string type + self.keep_client.put({}) class KeepPermissionTestCase(run_test_server.TestCaseWithServers): MAIN_SERVER = {} @@ -188,11 +204,12 @@ class KeepOptionalPermission(run_test_server.TestCaseWithServers): class KeepProxyTestCase(run_test_server.TestCaseWithServers): MAIN_SERVER = {} KEEP_SERVER = {} - KEEP_PROXY_SERVER = {'auth': 'admin'} + KEEP_PROXY_SERVER = {} @classmethod def setUpClass(cls): super(KeepProxyTestCase, cls).setUpClass() + run_test_server.authorize_with('active') cls.api_client = arvados.api('v1') def tearDown(self): @@ -231,44 +248,22 @@ class KeepProxyTestCase(run_test_server.TestCaseWithServers): self.assertTrue(keep_client.using_proxy) -class KeepClientServiceTestCase(unittest.TestCase): - def mock_keep_services(self, *services): - api_client = mock.MagicMock(name='api_client') - api_client.keep_services().accessible().execute.return_value = { - 'items_available': len(services), - 'items': [{ - 'uuid': 'zzzzz-bi6l4-{:015x}'.format(index), - 'owner_uuid': 'zzzzz-tpzed-000000000000000', - 'service_host': host, - 'service_port': port, - 'service_ssl_flag': ssl, - 'service_type': servtype, - } for index, (host, port, ssl, servtype) - in enumerate(services)], - } - return api_client - - def mock_n_keep_disks(self, service_count): - return self.mock_keep_services( - *[("keep0x{:x}".format(index), 80, False, 'disk') - for index in range(service_count)]) - - def get_service_roots(self, *services): - api_client = self.mock_keep_services(*services) +class KeepClientServiceTestCase(unittest.TestCase, tutil.ApiClientMock): + def get_service_roots(self, api_client): keep_client = arvados.KeepClient(api_client=api_client) - services = keep_client.weighted_service_roots('000000') + services = keep_client.weighted_service_roots(arvados.KeepLocator('0'*32)) return [urlparse.urlparse(url) for url in sorted(services)] def test_ssl_flag_respected_in_roots(self): - services = self.get_service_roots(('keep', 10, False, 'disk'), - ('keep', 20, True, 'disk')) - self.assertEqual(10, services[0].port) - self.assertEqual('http', services[0].scheme) - self.assertEqual(20, services[1].port) - self.assertEqual('https', services[1].scheme) + for ssl_flag in [False, True]: + services = self.get_service_roots(self.mock_keep_services( + service_ssl_flag=ssl_flag)) + self.assertEqual( + ('https' if ssl_flag else 'http'), services[0].scheme) def test_correct_ports_with_ipv6_addresses(self): - service = self.get_service_roots(('100::1', 10, True, 'proxy'))[0] + service = self.get_service_roots(self.mock_keep_services( + service_type='proxy', service_host='100::1', service_port=10, count=1))[0] self.assertEqual('100::1', service.hostname) self.assertEqual(10, service.port) @@ -277,58 +272,58 @@ class KeepClientServiceTestCase(unittest.TestCase): # when connected directly to a Keep server (i.e. non-proxy timeout) def test_get_timeout(self): - api_client = self.mock_keep_services(('keep', 10, False, 'disk')) - keep_client = arvados.KeepClient(api_client=api_client) + api_client = self.mock_keep_services(count=1) force_timeout = [socket.timeout("timed out")] - with mock.patch('requests.get', side_effect=force_timeout) as mock_request: + with tutil.mock_get(force_timeout) as mock_session: + keep_client = arvados.KeepClient(api_client=api_client) with self.assertRaises(arvados.errors.KeepReadError): keep_client.get('ffffffffffffffffffffffffffffffff') - self.assertTrue(mock_request.called) + self.assertTrue(mock_session.return_value.get.called) self.assertEqual( arvados.KeepClient.DEFAULT_TIMEOUT, - mock_request.call_args[1]['timeout']) + mock_session.return_value.get.call_args[1]['timeout']) def test_put_timeout(self): - api_client = self.mock_keep_services(('keep', 10, False, 'disk')) - keep_client = arvados.KeepClient(api_client=api_client) + api_client = self.mock_keep_services(count=1) force_timeout = [socket.timeout("timed out")] - with mock.patch('requests.put', side_effect=force_timeout) as mock_request: + with tutil.mock_put(force_timeout) as mock_session: + keep_client = arvados.KeepClient(api_client=api_client) with self.assertRaises(arvados.errors.KeepWriteError): keep_client.put('foo') - self.assertTrue(mock_request.called) + self.assertTrue(mock_session.return_value.put.called) self.assertEqual( arvados.KeepClient.DEFAULT_TIMEOUT, - mock_request.call_args[1]['timeout']) + mock_session.return_value.put.call_args[1]['timeout']) def test_proxy_get_timeout(self): # Force a timeout, verifying that the requests.get or # requests.put method was called with the proxy_timeout # setting rather than the default timeout. - api_client = self.mock_keep_services(('keep', 10, False, 'proxy')) - keep_client = arvados.KeepClient(api_client=api_client) + api_client = self.mock_keep_services(service_type='proxy', count=1) force_timeout = [socket.timeout("timed out")] - with mock.patch('requests.get', side_effect=force_timeout) as mock_request: + with tutil.mock_get(force_timeout) as mock_session: + keep_client = arvados.KeepClient(api_client=api_client) with self.assertRaises(arvados.errors.KeepReadError): keep_client.get('ffffffffffffffffffffffffffffffff') - self.assertTrue(mock_request.called) + self.assertTrue(mock_session.return_value.get.called) self.assertEqual( arvados.KeepClient.DEFAULT_PROXY_TIMEOUT, - mock_request.call_args[1]['timeout']) + mock_session.return_value.get.call_args[1]['timeout']) def test_proxy_put_timeout(self): # Force a timeout, verifying that the requests.get or # requests.put method was called with the proxy_timeout # setting rather than the default timeout. - api_client = self.mock_keep_services(('keep', 10, False, 'proxy')) - keep_client = arvados.KeepClient(api_client=api_client) + api_client = self.mock_keep_services(service_type='proxy', count=1) force_timeout = [socket.timeout("timed out")] - with mock.patch('requests.put', side_effect=force_timeout) as mock_request: + with tutil.mock_put(force_timeout) as mock_session: + keep_client = arvados.KeepClient(api_client=api_client) with self.assertRaises(arvados.errors.KeepWriteError): keep_client.put('foo') - self.assertTrue(mock_request.called) + self.assertTrue(mock_session.return_value.put.called) self.assertEqual( arvados.KeepClient.DEFAULT_PROXY_TIMEOUT, - mock_request.call_args[1]['timeout']) + mock_session.return_value.put.call_args[1]['timeout']) def test_probe_order_reference_set(self): # expected_order[i] is the probe order for @@ -346,10 +341,10 @@ class KeepClientServiceTestCase(unittest.TestCase): hashes = [ hashlib.md5("{:064x}".format(x)).hexdigest() for x in range(len(expected_order))] - api_client = self.mock_n_keep_disks(16) + api_client = self.mock_keep_services(count=16) keep_client = arvados.KeepClient(api_client=api_client) for i, hash in enumerate(hashes): - roots = keep_client.weighted_service_roots(hash) + roots = keep_client.weighted_service_roots(arvados.KeepLocator(hash)) got_order = [ re.search(r'//\[?keep0x([0-9a-f]+)', root).group(1) for root in roots] @@ -359,17 +354,17 @@ class KeepClientServiceTestCase(unittest.TestCase): hashes = [ hashlib.md5("{:064x}".format(x)).hexdigest() for x in range(100)] initial_services = 12 - api_client = self.mock_n_keep_disks(initial_services) + api_client = self.mock_keep_services(count=initial_services) keep_client = arvados.KeepClient(api_client=api_client) probes_before = [ - keep_client.weighted_service_roots(hash) for hash in hashes] + keep_client.weighted_service_roots(arvados.KeepLocator(hash)) for hash in hashes] for added_services in range(1, 12): - api_client = self.mock_n_keep_disks(initial_services+added_services) + api_client = self.mock_keep_services(count=initial_services+added_services) keep_client = arvados.KeepClient(api_client=api_client) total_penalty = 0 for hash_index in range(len(hashes)): probe_after = keep_client.weighted_service_roots( - hashes[hash_index]) + arvados.KeepLocator(hashes[hash_index])) penalty = probe_after.index(probes_before[hash_index][0]) self.assertLessEqual(penalty, added_services) total_penalty += penalty @@ -394,6 +389,132 @@ class KeepClientServiceTestCase(unittest.TestCase): min_penalty, max_penalty)) + def check_64_zeros_error_order(self, verb, exc_class): + data = '0' * 64 + if verb == 'get': + data = hashlib.md5(data).hexdigest() + '+1234' + # Arbitrary port number: + aport = random.randint(1024,65535) + api_client = self.mock_keep_services(service_port=aport, count=16) + keep_client = arvados.KeepClient(api_client=api_client) + with mock.patch('requests.' + verb, + side_effect=socket.timeout) as req_mock, \ + self.assertRaises(exc_class) as err_check: + getattr(keep_client, verb)(data) + urls = [urlparse.urlparse(url) + for url in err_check.exception.request_errors()] + self.assertEqual([('keep0x' + c, aport) for c in '3eab2d5fc9681074'], + [(url.hostname, url.port) for url in urls]) + + def test_get_error_shows_probe_order(self): + self.check_64_zeros_error_order('get', arvados.errors.KeepReadError) + + def test_put_error_shows_probe_order(self): + self.check_64_zeros_error_order('put', arvados.errors.KeepWriteError) + + def check_no_services_error(self, verb, exc_class): + api_client = mock.MagicMock(name='api_client') + api_client.keep_services().accessible().execute.side_effect = ( + arvados.errors.ApiError) + keep_client = arvados.KeepClient(api_client=api_client) + with self.assertRaises(exc_class) as err_check: + getattr(keep_client, verb)('d41d8cd98f00b204e9800998ecf8427e+0') + self.assertEqual(0, len(err_check.exception.request_errors())) + + def test_get_error_with_no_services(self): + self.check_no_services_error('get', arvados.errors.KeepReadError) + + def test_put_error_with_no_services(self): + self.check_no_services_error('put', arvados.errors.KeepWriteError) + + def check_errors_from_last_retry(self, verb, exc_class): + api_client = self.mock_keep_services(count=2) + req_mock = getattr(tutil, 'mock_{}_responses'.format(verb))( + "retry error reporting test", 500, 500, 403, 403) + with req_mock, tutil.skip_sleep, \ + self.assertRaises(exc_class) as err_check: + keep_client = arvados.KeepClient(api_client=api_client) + getattr(keep_client, verb)('d41d8cd98f00b204e9800998ecf8427e+0', + num_retries=3) + self.assertEqual([403, 403], [ + getattr(error, 'status_code', None) + for error in err_check.exception.request_errors().itervalues()]) + + def test_get_error_reflects_last_retry(self): + self.check_errors_from_last_retry('get', arvados.errors.KeepReadError) + + def test_put_error_reflects_last_retry(self): + self.check_errors_from_last_retry('put', arvados.errors.KeepWriteError) + + def test_put_error_does_not_include_successful_puts(self): + data = 'partial failure test' + data_loc = '{}+{}'.format(hashlib.md5(data).hexdigest(), len(data)) + api_client = self.mock_keep_services(count=3) + with tutil.mock_put_responses(data_loc, 200, 500, 500) as req_mock, \ + self.assertRaises(arvados.errors.KeepWriteError) as exc_check: + keep_client = arvados.KeepClient(api_client=api_client) + keep_client.put(data) + self.assertEqual(2, len(exc_check.exception.request_errors())) + + +class KeepClientGatewayTestCase(unittest.TestCase, tutil.ApiClientMock): + def mock_disks_and_gateways(self, disks=3, gateways=1): + self.gateways = [{ + 'uuid': 'zzzzz-bi6l4-gateway{:08d}'.format(i), + 'owner_uuid': 'zzzzz-tpzed-000000000000000', + 'service_host': 'gatewayhost{}'.format(i), + 'service_port': 12345, + 'service_ssl_flag': True, + 'service_type': 'gateway:test', + } for i in range(gateways)] + self.gateway_roots = [ + "https://[{service_host}]:{service_port}/".format(**gw) + for gw in self.gateways] + self.api_client = self.mock_keep_services( + count=disks, additional_services=self.gateways) + self.keepClient = arvados.KeepClient(api_client=self.api_client) + + @mock.patch('requests.Session') + def test_get_with_gateway_hint_first(self, MockSession): + MockSession.return_value.get.return_value = tutil.fake_requests_response( + code=200, body='foo', headers={'Content-Length': 3}) + self.mock_disks_and_gateways() + locator = 'acbd18db4cc2f85cedef654fccc4a4d8+3+K@' + self.gateways[0]['uuid'] + self.assertEqual('foo', self.keepClient.get(locator)) + self.assertEqual((self.gateway_roots[0]+locator,), + MockSession.return_value.get.call_args_list[0][0]) + + @mock.patch('requests.Session') + def test_get_with_gateway_hints_in_order(self, MockSession): + gateways = 4 + disks = 3 + MockSession.return_value.get.return_value = tutil.fake_requests_response( + code=404, body='') + self.mock_disks_and_gateways(gateways=gateways, disks=disks) + locator = '+'.join(['acbd18db4cc2f85cedef654fccc4a4d8+3'] + + ['K@'+gw['uuid'] for gw in self.gateways]) + with self.assertRaises(arvados.errors.NotFoundError): + self.keepClient.get(locator) + # Gateways are tried first, in the order given. + for i, root in enumerate(self.gateway_roots): + self.assertEqual((root+locator,), + MockSession.return_value.get.call_args_list[i][0]) + # Disk services are tried next. + for i in range(gateways, gateways+disks): + self.assertRegexpMatches( + MockSession.return_value.get.call_args_list[i][0][0], + r'keep0x') + + @mock.patch('requests.Session') + def test_get_with_remote_proxy_hint(self, MockSession): + MockSession.return_value.get.return_value = tutil.fake_requests_response( + code=200, body='foo', headers={'Content-Length': 3}) + self.mock_disks_and_gateways() + locator = 'acbd18db4cc2f85cedef654fccc4a4d8+3+K@xyzzy' + self.assertEqual('foo', self.keepClient.get(locator)) + self.assertEqual(('https://keep.xyzzy.arvadosapi.com/'+locator,), + MockSession.return_value.get.call_args_list[0][0]) + class KeepClientRetryTestMixin(object): # Testing with a local Keep store won't exercise the retry behavior. @@ -494,17 +615,13 @@ class KeepClientRetryGetTestCase(KeepClientRetryTestMixin, unittest.TestCase): self.check_success(locator=self.HINTED_LOCATOR) def test_try_next_server_after_timeout(self): - side_effects = [ - socket.timeout("timed out"), - tutil.fake_requests_response(200, self.DEFAULT_EXPECT)] - with mock.patch('requests.get', - side_effect=iter(side_effects)): + with tutil.mock_get([ + socket.timeout("timed out"), + tutil.fake_requests_response(200, self.DEFAULT_EXPECT)]): self.check_success(locator=self.HINTED_LOCATOR) def test_retry_data_with_wrong_checksum(self): - side_effects = (tutil.fake_requests_response(200, s) - for s in ['baddata', self.TEST_DATA]) - with mock.patch('requests.get', side_effect=side_effects): + with tutil.mock_get((tutil.fake_requests_response(200, s) for s in ['baddata', self.TEST_DATA])): self.check_success(locator=self.HINTED_LOCATOR)