21 global_client_object = None
29 def global_client_object():
30 global global_client_object
31 if global_client_object == None:
32 global_client_object = KeepClient()
33 return global_client_object
36 def get(locator, **kwargs):
37 return Keep.global_client_object().get(locator, **kwargs)
40 def put(data, **kwargs):
41 return Keep.global_client_object().put(data, **kwargs)
43 class KeepClient(object):
45 class ThreadLimiter(object):
47 Limit the number of threads running at a given time to
48 {desired successes} minus {successes reported}. When successes
49 reported == desired, wake up the remaining threads and tell
52 Should be used in a "with" block.
54 def __init__(self, todo):
57 self._todo_lock = threading.Semaphore(todo)
58 self._done_lock = threading.Lock()
61 self._todo_lock.acquire()
64 def __exit__(self, type, value, traceback):
65 self._todo_lock.release()
67 def shall_i_proceed(self):
69 Return true if the current thread should do stuff. Return
70 false if the current thread should just stop.
73 return (self._done < self._todo)
75 def increment_done(self):
77 Report that the current thread was successful.
84 Return how many successes were reported.
89 class KeepWriterThread(threading.Thread):
91 Write a blob of data to the given Keep server. Call
92 increment_done() of the given ThreadLimiter if the write
95 def __init__(self, **kwargs):
96 super(KeepClient.KeepWriterThread, self).__init__()
100 with self.args['thread_limiter'] as limiter:
101 if not limiter.shall_i_proceed():
102 # My turn arrived, but the job has been done without
105 logging.debug("KeepWriterThread %s proceeding %s %s" %
106 (str(threading.current_thread()),
107 self.args['data_hash'],
108 self.args['service_root']))
110 url = self.args['service_root'] + self.args['data_hash']
111 api_token = config.get('ARVADOS_API_TOKEN')
112 headers = {'Authorization': "OAuth2 %s" % api_token}
114 resp, content = h.request(url.encode('utf-8'), 'PUT',
116 body=self.args['data'])
117 if (resp['status'] == '401' and
118 re.match(r'Timestamp verification failed', content)):
119 body = KeepClient.sign_for_old_server(
120 self.args['data_hash'],
123 resp, content = h.request(url.encode('utf-8'), 'PUT',
126 if re.match(r'^2\d\d$', resp['status']):
127 logging.debug("KeepWriterThread %s succeeded %s %s" %
128 (str(threading.current_thread()),
129 self.args['data_hash'],
130 self.args['service_root']))
131 return limiter.increment_done()
132 logging.warning("Request fail: PUT %s => %s %s" %
133 (url, resp['status'], content))
134 except (httplib2.HttpLib2Error, httplib.HTTPException) as e:
135 logging.warning("Request fail: PUT %s => %s: %s" %
136 (url, type(e), str(e)))
139 self.lock = threading.Lock()
140 self.service_roots = None
142 def shuffled_service_roots(self, hash):
143 if self.service_roots == None:
145 keep_disks = api().keep_disks().list().execute()['items']
146 roots = (("http%s://%s:%d/" %
147 ('s' if f['service_ssl_flag'] else '',
151 self.service_roots = sorted(set(roots))
152 logging.debug(str(self.service_roots))
155 pool = self.service_roots[:]
159 if len(pseq) < len(hash) / 4: # first time around
160 seed = hash[-4:] + hash
163 probe = int(seed[0:8], 16) % len(pool)
164 pseq += [pool[probe]]
165 pool = pool[:probe] + pool[probe+1:]
167 logging.debug(str(pseq))
170 def get(self, locator):
171 if re.search(r',', locator):
172 return ''.join(self.get(x) for x in locator.split(','))
173 if 'KEEP_LOCAL_STORE' in os.environ:
174 return KeepClient.local_store_get(locator)
175 expect_hash = re.sub(r'\+.*', '', locator)
176 for service_root in self.shuffled_service_roots(expect_hash):
177 url = service_root + expect_hash
178 api_token = config.get('ARVADOS_API_TOKEN')
179 headers = {'Authorization': "OAuth2 %s" % api_token,
180 'Accept': 'application/octet-stream'}
181 blob = self.get_url(url, headers, expect_hash)
184 for location_hint in re.finditer(r'\+K@([a-z0-9]+)', locator):
185 instance = location_hint.group(1)
186 url = 'http://keep.' + instance + '.arvadosapi.com/' + expect_hash
187 blob = self.get_url(url, {}, expect_hash)
190 raise arvados.errors.NotFoundError("Block not found: %s" % expect_hash)
192 def get_url(self, url, headers, expect_hash):
195 resp, content = h.request(url.encode('utf-8'), 'GET',
197 if re.match(r'^2\d\d$', resp['status']):
198 m = hashlib.new('md5')
201 if md5 == expect_hash:
203 logging.warning("Checksum fail: md5(%s) = %s" % (url, md5))
204 except Exception as e:
205 logging.info("Request fail: GET %s => %s: %s" %
206 (url, type(e), str(e)))
209 def put(self, data, **kwargs):
210 if 'KEEP_LOCAL_STORE' in os.environ:
211 return KeepClient.local_store_put(data)
212 m = hashlib.new('md5')
214 data_hash = m.hexdigest()
216 want_copies = kwargs.get('copies', 2)
217 if not (want_copies > 0):
220 thread_limiter = KeepClient.ThreadLimiter(want_copies)
221 for service_root in self.shuffled_service_roots(data_hash):
222 t = KeepClient.KeepWriterThread(data=data,
224 service_root=service_root,
225 thread_limiter=thread_limiter)
230 have_copies = thread_limiter.done()
231 if have_copies == want_copies:
232 return (data_hash + '+' + str(len(data)))
233 raise arvados.errors.KeepWriteError(
234 "Write fail for %s: wanted %d but wrote %d" %
235 (data_hash, want_copies, have_copies))
238 def sign_for_old_server(data_hash, data):
239 return (("-----BEGIN PGP SIGNED MESSAGE-----\n\n\n%d %s\n-----BEGIN PGP SIGNATURE-----\n\n-----END PGP SIGNATURE-----\n" % (int(time.time()), data_hash)) + data)
243 def local_store_put(data):
244 m = hashlib.new('md5')
247 locator = '%s+%d' % (md5, len(data))
248 with open(os.path.join(os.environ['KEEP_LOCAL_STORE'], md5 + '.tmp'), 'w') as f:
250 os.rename(os.path.join(os.environ['KEEP_LOCAL_STORE'], md5 + '.tmp'),
251 os.path.join(os.environ['KEEP_LOCAL_STORE'], md5))
255 def local_store_get(locator):
256 r = re.search('^([0-9a-f]{32,})', locator)
258 raise arvados.errors.NotFoundError(
259 "Invalid data locator: '%s'" % locator)
260 if r.group(0) == config.EMPTY_BLOCK_LOCATOR.split('+')[0]:
262 with open(os.path.join(os.environ['KEEP_LOCAL_STORE'], r.group(0)), 'r') as f: