def inner_func(q: eventlet.Queue):
a = MySingleton()
b = MySingleton()
q.put((
a.uuid,
b.uuid,
))
class Dispatcher(EventThread):
def __init__(self):
self.proposer_queue = Queue()
self.acceptor_queue = Queue()
def _run(self):
while True:
#get recved string
u = do_recv()
m = Message()
if not m.depack(u):
log.debug('depack %s error', u)
continue
#check timestamp
if m.timestamp - get_utc_time() > conf.proposer_timeout:
log.debug('receive expired package, drop it')
continue
#check version
if m.version != VERSION:
log.debug('version not match')
continue
#check signature
if m.signature != m.build_hmac_signature():
log.info('message signature failed, may changed in network')
continue
if m.method in (PREPARE, ACCEPT, LEARN, RENEW):
self.acceptor_queue.put(m)
elif m.method in (NACK, ACK, OUTDATE):
self.proposer_queue.put(m)
def subscribe(self, q: eventlet.Queue) -> NoReturn:
"""Green thread process waiting for NOTIFYs on the channel and feeding
them to the queue.
Args:
q: event queue through which to pipe NOTIFY events to the main
thread.
"""
# Subscribe to notification channel
self.database.listen_on_channel(self.channel, self.key)
# Infinite listening loop
while True:
# self.database.pool._used[self.key] is the connection object
# corresponding to [key] in the conneciton pool
# spawns a green thread and return control once there is a
# notification to read
trampoline(self.database.pool._used[self.key], read=True)
# once there is a notification --> poll
self.database.pool._used[self.key].poll()
while self.database.pool._used[self.key].notifies:
# extract notify:
notify = self.database.pool._used[self.key].notifies.pop()
# block until slot available in queue to insert Notify:
q.put(notify)
class WSTestGenerator(WebSocketView):
def handle_websocket(self, ws):
self._ws = ws
return super(RangeWebsocket, self)
def handler(self, ws):
self.queue = Queue()
while True:
m = ws.wait()
# import ipdb; ipdb.set_trace()
if m is None:
break
self.queue.put(m)
class WSTestGenerator(WebSocketView):
def handle_websocket(self, ws):
self._ws = ws
return super(RangeWebsocket, self)
def handler(self, ws):
self.queue = Queue()
while True:
m = ws.wait()
# import ipdb; ipdb.set_trace()
if m is None:
break
self.queue.put(m)
def take_action(self, parsed_args):
self.log.debug('take_action(%s)', parsed_args)
digits = self.app.client_manager.get_meta1_digits()
workers_count = parsed_args.workers
conf = {'namespace': self.app.client_manager.namespace}
if parsed_args.proxy:
conf.update({'proxyd_url': parsed_args.proxy})
else:
ns_conf = load_namespace_conf(conf['namespace'])
proxy = ns_conf.get('proxy')
conf.update({'proxyd_url': proxy})
workers = list()
with green.ContextPool(workers_count) as pool:
pile = GreenPile(pool)
prefix_queue = Queue(16)
# Prepare some workers
for i in range(workers_count):
w = WarmupWorker(conf, self.log)
workers.append(w)
pile.spawn(w.run, prefix_queue)
# Feed the queue
trace_increment = 0.01
trace_next = trace_increment
sent, total = 0, float(count_prefixes(digits))
for prefix in generate_prefixes(digits):
sent += 1
prefix_queue.put(prefix)
# Display the progression
ratio = float(sent) / total
if ratio >= trace_next:
self.log.info("... %d%%", int(ratio * 100.0))
trace_next += trace_increment
self.log.debug("Send the termination marker")
prefix_queue.join()
self.log.info("All the workers are done")
def test_connection_pooling(self):
with patch('swift.common.memcached.socket') as mock_module:
# patch socket, stub socket.socket, mock sock
mock_sock = mock_module.socket.return_value
# track clients waiting for connections
connected = []
connections = Queue()
def wait_connect(addr):
connected.append(addr)
connections.get()
mock_sock.connect = wait_connect
memcache_client = memcached.MemcacheRing(['1.2.3.4:11211'],
connect_timeout=10)
# sanity
self.assertEquals(1, len(memcache_client._client_cache))
for server, pool in memcache_client._client_cache.items():
self.assertEquals(2, pool.max_size)
# make 10 requests "at the same time"
p = GreenPool()
for i in range(10):
p.spawn(memcache_client.set, 'key', 'value')
for i in range(3):
sleep(0.1)
self.assertEquals(2, len(connected))
# give out a connection
connections.put(None)
for i in range(3):
sleep(0.1)
self.assertEquals(2, len(connected))
# finish up
for i in range(8):
connections.put(None)
self.assertEquals(2, len(connected))
p.waitall()
self.assertEquals(2, len(connected))
def test_connection_pooling(self):
with patch('swift.common.memcached.socket') as mock_module:
# patch socket, stub socket.socket, mock sock
mock_sock = mock_module.socket.return_value
# track clients waiting for connections
connected = []
connections = Queue()
errors = []
def wait_connect(addr):
connected.append(addr)
sleep(0.1) # yield
val = connections.get()
if val is not None:
errors.append(val)
mock_sock.connect = wait_connect
memcache_client = memcached.MemcacheRing(['1.2.3.4:11211'],
connect_timeout=10)
# sanity
self.assertEquals(1, len(memcache_client._client_cache))
for server, pool in memcache_client._client_cache.items():
self.assertEqual(2, pool.max_size)
# make 10 requests "at the same time"
p = GreenPool()
for i in range(10):
p.spawn(memcache_client.set, 'key', 'value')
for i in range(3):
sleep(0.1)
self.assertEqual(2, len(connected))
# give out a connection
connections.put(None)
# at this point, only one connection should have actually been
# created, the other is in the creation step, and the rest of the
# clients are not attempting to connect. we let this play out a
# bit to verify.
for i in range(3):
sleep(0.1)
self.assertEqual(2, len(connected))
# finish up, this allows the final connection to be created, so
# that all the other clients can use the two existing connections
# and no others will be created.
connections.put(None)
connections.put('nono')
self.assertEqual(2, len(connected))
p.waitall()
self.assertEqual(2, len(connected))
self.assertEqual(0, len(errors),
"A client was allowed a third connection")
connections.get_nowait()
self.assertTrue(connections.empty())
def test_connection_pooling(self):
with patch('swift.common.memcached.socket') as mock_module:
# patch socket, stub socket.socket, mock sock
mock_sock = mock_module.socket.return_value
# track clients waiting for connections
connected = []
connections = Queue()
errors = []
def wait_connect(addr):
connected.append(addr)
sleep(0.1) # yield
val = connections.get()
if val is not None:
errors.append(val)
mock_sock.connect = wait_connect
memcache_client = memcached.MemcacheRing(['1.2.3.4:11211'],
connect_timeout=10)
# sanity
self.assertEquals(1, len(memcache_client._client_cache))
for server, pool in memcache_client._client_cache.items():
self.assertEqual(2, pool.max_size)
# make 10 requests "at the same time"
p = GreenPool()
for i in range(10):
p.spawn(memcache_client.set, 'key', 'value')
for i in range(3):
sleep(0.1)
self.assertEqual(2, len(connected))
# give out a connection
connections.put(None)
# at this point, only one connection should have actually been
# created, the other is in the creation step, and the rest of the
# clients are not attempting to connect. we let this play out a
# bit to verify.
for i in range(3):
sleep(0.1)
self.assertEqual(2, len(connected))
# finish up, this allows the final connection to be created, so
# that all the other clients can use the two existing connections
# and no others will be created.
connections.put(None)
connections.put('nono')
self.assertEqual(2, len(connected))
p.waitall()
self.assertEqual(2, len(connected))
self.assertEqual(0, len(errors),
"A client was allowed a third connection")
connections.get_nowait()
self.assertTrue(connections.empty())
class GreenletsThread(Thread):
"""
Main thread for the program. If running stand alone this will be running
as a greenlet instead.
"""
def __init__ (self, server_url, login_params):
self.running = True
self.agent = True
self.cmd_out_queue = []
self.cmd_in_queue = []
self.out_queue = Queue()
self.in_queue = Queue()
self.server_url = server_url
self.login_params = login_params
Thread.__init__(self)
def apply_position(self, obj_uuid, pos, rot=None):
cmd = ['pos', obj_uuid, pos, rot]
self.addCmd(cmd)
def __getattr__(self, name):
return ProxyFunction(name, self)
def apply_scale(self, obj_uuid, scale):
cmd = ['scale', obj_uuid, scale]
self.addCmd(cmd)
def run(self):
agent = AgentManager(self.in_queue,
self.out_queue)
error = agent.login(self.server_url, self.login_params)
if error:
self.out_queue.put(["error", str(error)])
self.out_queue.put(["agentquit", str(error)])
while self.out_queue.qsize():
api.sleep(0.1)
agent.logger.debug("Quitting")
self.agent = agent
self.running = False
def addCmd(self, cmd):
self.in_queue.put(cmd)
def getQueue(self):
out_queue = []
while self.out_queue.qsize():
out_queue.append(self.out_queue.get())
return out_queue
class GreenletsThread(Thread):
"""
Main thread for the program. If running stand alone this will be running
as a greenlet instead.
"""
def __init__(self, server_url, login_params):
self.running = True
self.agent = True
self.cmd_out_queue = []
self.cmd_in_queue = []
self.out_queue = Queue()
self.in_queue = Queue()
self.server_url = server_url
self.login_params = login_params
Thread.__init__(self)
def apply_position(self, obj_uuid, pos, rot=None):
cmd = ['pos', obj_uuid, pos, rot]
self.addCmd(cmd)
def __getattr__(self, name):
return ProxyFunction(name, self)
def apply_scale(self, obj_uuid, scale):
cmd = ['scale', obj_uuid, scale]
self.addCmd(cmd)
def run(self):
agent = AgentManager(self.in_queue, self.out_queue)
error = agent.login(self.server_url, self.login_params)
if error:
self.out_queue.put(["error", str(error)])
self.out_queue.put(["agentquit", str(error)])
while self.out_queue.qsize():
api.sleep(0.1)
agent.logger.debug("Quitting")
self.agent = agent
self.running = False
def addCmd(self, cmd):
self.in_queue.put(cmd)
def getQueue(self):
out_queue = []
while self.out_queue.qsize():
out_queue.append(self.out_queue.get())
return out_queue
def run_stock_parser():
symbol_q = Queue()
price_q = Queue()
stock_symbols = []
with open('symbols.txt', 'r') as symfile:
for n, line in enumerate(symfile):
sym = line.strip()
if sym:
stock_symbols.append(sym)
ncpu = cpu_count()
pool = [spawn(read_stock_worker, symbol_q, price_q) for _ in range(ncpu * 2)]
output = spawn(write_output_file, price_q)
for symbol in stock_symbols:
symbol_q.put(symbol)
symbol_q.put(_sentinel)
for p in pool:
p.wait()
price_q.put(_sentinel)
output.wait()
class ECWriter(object):
"""
Writes an EC chunk
"""
def __init__(self, chunk, conn):
self._chunk = chunk
self._conn = conn
self.failed = False
self.bytes_transferred = 0
self.checksum = hashlib.md5()
@property
def chunk(self):
return self._chunk
@property
def conn(self):
return self._conn
@classmethod
def connect(cls, chunk, sysmeta, reqid=None):
raw_url = chunk["url"]
parsed = urlparse(raw_url)
chunk_path = parsed.path.split('/')[-1]
h = {}
h["transfer-encoding"] = "chunked"
h[chunk_headers["content_id"]] = sysmeta['id']
h[chunk_headers["content_path"]] = sysmeta['content_path']
h[chunk_headers["content_chunkmethod"]] = sysmeta['chunk_method']
h[chunk_headers["container_id"]] = sysmeta['container_id']
h[chunk_headers["chunk_pos"]] = chunk["pos"]
h[chunk_headers["chunk_id"]] = chunk_path
h[chunk_headers["content_policy"]] = sysmeta['policy']
h[chunk_headers["content_version"]] = sysmeta['version']
if reqid:
h['X-oio-req-id'] = reqid
# in the trailer
# metachunk_size & metachunk_hash
h["Trailer"] = (chunk_headers["metachunk_size"],
chunk_headers["metachunk_hash"])
with ConnectionTimeout(io.CONNECTION_TIMEOUT):
conn = io.http_connect(
parsed.netloc, 'PUT', parsed.path, h)
conn.chunk = chunk
return cls(chunk, conn)
def start(self, pool):
# we use eventlet Queue to pass data to the send coroutine
self.queue = Queue(io.PUT_QUEUE_DEPTH)
# spawn the send coroutine
pool.spawn(self._send)
def _send(self):
# this is the send coroutine loop
while True:
# fetch input data from the queue
d = self.queue.get()
# use HTTP transfer encoding chunked
# to write data to RAWX
if not self.failed:
# format the chunk
to_send = "%x\r\n%s\r\n" % (len(d), d)
try:
with ChunkWriteTimeout(io.CHUNK_TIMEOUT):
self.conn.send(to_send)
self.bytes_transferred += len(d)
except (Exception, ChunkWriteTimeout) as e:
self.failed = True
msg = str(e)
logger.warn("Failed to write to %s (%s)", self.chunk, msg)
self.chunk['error'] = msg
self.queue.task_done()
def wait(self):
# wait until all data in the queue
# has been processed by the send coroutine
if self.queue.unfinished_tasks:
self.queue.join()
def send(self, data):
# do not send empty data because
# this will end the chunked body
if not data:
return
# put the data to send into the queue
# it will be processed by the send coroutine
self.queue.put(data)
def finish(self, metachunk_size, metachunk_hash):
parts = [
'0\r\n',
'%s: %s\r\n' % (chunk_headers['metachunk_size'],
metachunk_size),
'%s: %s\r\n' % (chunk_headers['metachunk_hash'],
metachunk_hash),
'\r\n'
]
to_send = "".join(parts)
self.conn.send(to_send)
def getresponse(self):
# read the HTTP response from the connection
with Timeout(io.CHUNK_TIMEOUT):
self.resp = self.conn.getresponse()
return self.resp
class EcChunkWriter(object):
"""
Writes an EC chunk
"""
def __init__(self, chunk, conn, write_timeout=None, **_kwargs):
self._chunk = chunk
self._conn = conn
self.failed = False
self.bytes_transferred = 0
self.checksum = hashlib.md5()
self.write_timeout = write_timeout or io.CHUNK_TIMEOUT
# we use eventlet Queue to pass data to the send coroutine
self.queue = Queue(io.PUT_QUEUE_DEPTH)
@property
def chunk(self):
return self._chunk
@property
def conn(self):
return self._conn
@classmethod
def connect(cls,
chunk,
sysmeta,
reqid=None,
connection_timeout=None,
write_timeout=None,
**_kwargs):
raw_url = chunk["url"]
parsed = urlparse(raw_url)
chunk_path = parsed.path.split('/')[-1]
hdrs = headers_from_object_metadata(sysmeta)
if reqid:
hdrs['X-oio-req-id'] = reqid
hdrs[chunk_headers["chunk_pos"]] = chunk["pos"]
hdrs[chunk_headers["chunk_id"]] = chunk_path
# in the trailer
# metachunk_size & metachunk_hash
hdrs["Trailer"] = ', '.join(
(chunk_headers["metachunk_size"], chunk_headers["metachunk_hash"],
chunk_headers["chunk_hash"]))
with green.ConnectionTimeout(connection_timeout
or io.CONNECTION_TIMEOUT):
conn = io.http_connect(parsed.netloc, 'PUT', parsed.path, hdrs)
conn.chunk = chunk
return cls(chunk, conn, write_timeout=write_timeout)
def start(self, pool):
"""Spawn the send coroutine"""
pool.spawn(self._send)
def _send(self):
"""Send coroutine loop"""
while True:
# fetch input data from the queue
data = self.queue.get()
# use HTTP transfer encoding chunked
# to write data to RAWX
if not self.failed:
# format the chunk
to_send = "%x\r\n%s\r\n" % (len(data), data)
try:
with green.ChunkWriteTimeout(self.write_timeout):
self.conn.send(to_send)
self.bytes_transferred += len(data)
except (Exception, green.ChunkWriteTimeout) as exc:
self.failed = True
msg = str(exc)
logger.warn("Failed to write to %s (%s)", self.chunk, msg)
self.chunk['error'] = 'write: %s' % msg
self.queue.task_done()
def wait(self):
"""
Wait until all data in the queue
has been processed by the send coroutine
"""
if self.queue.unfinished_tasks:
self.queue.join()
def send(self, data):
# do not send empty data because
# this will end the chunked body
if not data:
return
# put the data to send into the queue
# it will be processed by the send coroutine
self.queue.put(data)
def finish(self, metachunk_size, metachunk_hash):
"""Send metachunk_size and metachunk_hash as trailers"""
parts = [
'0\r\n',
'%s: %s\r\n' % (chunk_headers['metachunk_size'], metachunk_size),
'%s: %s\r\n' % (chunk_headers['metachunk_hash'], metachunk_hash),
'%s: %s\r\n' %
(chunk_headers['chunk_hash'], self.checksum.hexdigest()), '\r\n'
]
to_send = "".join(parts)
self.conn.send(to_send)
def getresponse(self):
"""Read the HTTP response from the connection"""
# As the server may buffer data before writing it to non-volatile
# storage, we don't know if we have to wait while sending data or
# while reading response, thus we apply the same timeout to both.
with Timeout(self.write_timeout):
return self.conn.getresponse()
class Manager(object):
"""Class encapsulating Heroshi URL server state."""
def __init__(self):
self.active = False
self.prefetch_queue = Queue(settings.prefetch['queue_size'])
self.prefetch_thread = spawn(self.prefetch_worker)
self.prefetch_thread.link(reraise_errors, greenthread.getcurrent())
self.given_items = Cache()
self.postreport_queue = Queue(settings.postreport['queue_size'])
self.postreport_thread = spawn(self.postreport_worker)
self.postreport_thread.link(reraise_errors, greenthread.getcurrent())
self.storage_connections = eventlet.pools.Pool(max_size=settings.storage['max_connections'])
self.storage_connections.create = StorageConnection
def close(self):
self.active = False
self.prefetch_thread.kill()
self.postreport_thread.kill()
def ping_storage(self):
with self.storage_connections.item() as storage:
pass
def get_from_prefetch_queue(self, size):
result = []
while len(result) < size:
sleep()
try:
pack = self.prefetch_queue.get(timeout=settings.prefetch['get_timeout'])
except eventlet.queue.Empty:
break
result.extend(pack)
return result
def prefetch_worker(self):
if not self.active:
sleep(0.01)
while self.active:
with self.storage_connections.item() as storage:
docs = storage.query_new_random(settings.prefetch['single_limit'])
if len(docs) == 0:
sleep(10.)
continue
else:
# Note: putting a *list* as a single item into queue
self.prefetch_queue.put(docs)
# and respawn again
self.prefetch_thread = spawn(self.prefetch_worker)
@log_exceptions
def _postreport_worker(self):
docs = []
while len(docs) < settings.postreport['flush_size']: # inner accumulator loop
try:
item = self.postreport_queue.get(timeout=settings.postreport['flush_delay'])
except eventlet.queue.Empty:
break
# Quick dirty duplicate filtering.
# Note that this code only finds dups in current "flush pack". `report_result` uses
# `is_duplicate_report` which finds dups in whole `postreport_queue` but it can't find dups here.
# Thus two dups searchers.
# It is still possible that at most 2 duplicate reports exist: one in `postreport_queue`
# and one in current "flush pack". This is acceptable, because most of the dups are filtered out.
for doc in docs:
if item['url'] == doc['url']:
item = None
break
if item is None:
continue
if 'result' not in item:
# It's a link, found on some reported page.
# Just add it to bulk insert, don't try to update any document here.
docs.append(item)
continue
docs.append(item)
if not docs:
return
with self.storage_connections.item() as storage:
for doc in docs:
content = doc.pop('content', None)
storage.save(doc)
if content is None:
continue
headers = doc.get('headers') or {}
content_type = headers.get('content-type', "application/octet-stream")
storage.save_content(doc, content, content_type)
def postreport_worker(self):
if not self.active:
sleep(0.01)
while self.active:
self._postreport_worker()
# and respawn again
self.postreport_thread = spawn(self.prefetch_worker)
@log_exceptions
def crawl_queue(self, request):
limit = max(int(request.POST['limit']), settings.api['max_queue_limit'])
time_now = datetime.datetime.now()
doc_list = self.get_from_prefetch_queue(limit)
for doc in doc_list:
if isinstance(doc['visited'], basestring):
doc['visited'] = datetime.datetime.strptime(doc['visited'], TIME_FORMAT)
self.given_items.set(doc['url'], doc, settings.prefetch['cache_timeout'])
def is_old(doc):
"""Predicate tells if page was never visited or visited long enough ago.
Worker SHOULD NOT visit URI, if this function returns False.
"""
if doc['visited'] is None:
return True
diff = time_now - doc['visited']
return diff > datetime.timedelta(minutes=settings.api['min_revisit_minutes'])
doc_list = filter(is_old, doc_list)
def make_queue_item(doc):
if isinstance(doc['visited'], datetime.datetime):
doc['visited'] = doc['visited'].strftime(TIME_FORMAT)
filter_fields = ('url', 'headers', 'visited',)
return dict( (k,v) for (k,v) in doc.iteritems() if k in filter_fields )
queue = map(make_queue_item, doc_list)
return queue
def is_duplicate_report(self, url):
"""Quick dirty duplicate searching."""
for doc in self.postreport_queue.queue:
if url == doc['url']:
return True
return False
def force_append_links(self, links):
# 1. remove duplicates
links = set(links)
# 2. put links into queue
for url in links:
new_doc = {'url': url, 'parent': None, 'visited': None}
self.postreport_queue.put(new_doc)
@log_exceptions
def report_result(self, request):
report = json.loads(request.body)
# `report['links']` now used only to force insertion of new URLs into
# Heroshi crawling queue via bin/heroshi-append script.
# So, if a more sophisticated way to append new URLs is to arise,
# remove this code.
if report['url'] is None:
self.force_append_links(report['links'])
return
if self.is_duplicate_report(report['url']):
return
# accept report into postreport_queue for later persistent saving
try:
doc = self.given_items[report['url']]
except KeyError:
self.postreport_queue.put(report)
else:
doc.update(report)
self.postreport_queue.put(doc)
return None
def main():
args = options()
global ACCOUNT, PROXY, QUEUE, NS, VERBOSE, TIMEOUT
global COUNTERS, ELECTIONS
ACCOUNT = args.account
NS = args.namespace
VERBOSE = args.verbose
TIMEOUT = args.timeout
PROXY = ObjectStorageApi(NS)
ELECTIONS = AtomicInteger()
num_worker_threads = int(args.max_worker)
print("Using %d workers" % num_worker_threads)
total_objects = {'size': 0, 'files': 0, 'elapsed': 0}
total_containers = {'size': 0, 'files': 0, 'elapsed': 0}
for path in args.path:
path = path.rstrip('/')
if '/' in path:
bucket, path = path.split('/', 1)
else:
bucket = path
path = ""
containers = []
QUEUE = Queue()
pool = eventlet.GreenPool(num_worker_threads)
for i in range(num_worker_threads):
pool.spawn(worker_objects)
COUNTERS = AtomicInteger()
_bucket = container_hierarchy(bucket, path)
# we don't use placeholders, we use prefix path as prefix
for entry in full_list(prefix=container_hierarchy(bucket, path)):
name, _files, _size, _ = entry
if name != _bucket and not name.startswith(_bucket + '%2F'):
continue
if _files:
QUEUE.put(name)
containers.append(name)
# we have to wait all objects
print("Waiting flush of objects")
report = args.report
while not QUEUE.empty():
ts = time.time()
while time.time() - ts < report and not QUEUE.empty():
time.sleep(1)
diff = time.time() - ts
val = COUNTERS.reset()
elections = ELECTIONS.reset()
print("Objects: %5.2f / Size: %5.2f" %
(val[0] / diff, val[1] / diff),
"Elections failed: %5.2f/s total: %d" %
(elections[0] / diff, ELECTIONS.total()[0]),
" " * 20,
end='\r')
sys.stdout.flush()
print("Waiting end of workers")
QUEUE.join()
val = COUNTERS.total()
total_objects['files'] += val[0]
total_objects['size'] += val[1]
total_objects['elapsed'] += COUNTERS.time()
COUNTERS = AtomicInteger()
QUEUE = Queue()
for i in range(num_worker_threads):
pool.spawn(worker_container)
print("We have to delete", len(containers), "containers")
for container in containers:
QUEUE.put(container)
while not QUEUE.empty():
ts = time.time()
while time.time() - ts < report and not QUEUE.empty():
time.sleep(1)
diff = time.time() - ts
val = COUNTERS.reset()
elections = ELECTIONS.reset()
print("Containers: %5.2f" % (val[0] / diff),
"Elections failed: %5.2f/s total: %d" %
(elections[0] / diff, ELECTIONS.total()[0]),
" " * 20,
end='\r')
sys.stdout.flush()
QUEUE.join()
val = COUNTERS.total()
total_containers['files'] += val[0]
total_containers['size'] += val[1]
total_containers['elapsed'] += COUNTERS.time()
print("""
Objects:
- ran during {o[elapsed]:5.2f}
- {o[files]} objects removed (size {size})
- {o_file_avg:5.2f} objects/s ({o_size_avg} avg. size/s)
""".format(o=total_objects,
size=show(total_objects['size'], True),
o_file_avg=total_objects['files'] / total_objects['elapsed'],
o_size_avg=show(total_objects['size'] / total_objects['elapsed'],
True)))
print("""
Containers:
- ran during {o[elapsed]:5.2f}
- {o[files]} containers
- {o_file_avg:5.2f} containers/s
""".format(o=total_containers,
o_file_avg=total_containers['files'] / total_containers['elapsed']))
print("Elections failed: %d" % ELECTIONS.total()[0])
class Crawler(object):
"""
A crawler will traverse all the pages of a site and process the content
in a defined way.
:param init_urls: the very first urls to start with.
:param q: the queue that stores all urls to be crawled
:param urls: a set stores all urls already crawled
"""
def __init__(self, init_urls, max_workers=200):
self.init_urls = init_urls
self.max_workers = max_workers
self.q = Queue()
self.urls = set()
self.s = requests.Session()
self.root_hosts = set()
for url in init_urls:
self.q.put(url)
self.urls.add(url)
self.root_hosts.add(get_netloc(url))
def url_allowed(self, url):
"""Check if given url will be crawled.
Current, only if the url belongs to the same host as init_urls.
"""
return get_netloc(url) in self.root_hosts
def save(self, response):
"""Save data at the given url."""
raise NotImplementedError(
"Please implement your own save logic in subclass.")
def parse(self, response):
self.save(response)
new_links = set()
for url in self.find_links(response):
if url not in self.urls and self.url_allowed(url):
new_links.add(url)
self.urls.add(url)
self.q.put(url)
if len(new_links) != 0:
print("Find %d new urls to crawl" % len(new_links))
def fetch(self, url):
"""Fetch content of the url from network."""
response = self.s.get(url)
print("Getting content from %s, length: %d" % (url,
len(response.content)))
return response
def work(self, i):
"""Define the work process.
Retrieve a url from queue, fetch the content from it,
process it and get new urls to crawl.
Continue the process until all pages are crawled.
:param i: indicate the worker number
"""
while True:
url = self.q.get()
print("Worker %d: Getting url %s from queue." % (i, url))
response = self.fetch(url)
self.parse(response)
self.q.task_done()
def run(self):
"""Start the crawling process.
This is the main entrance for our crawler. It will start several
workers, crawling in parallel.
"""
pool = eventlet.GreenPool()
start = time.time()
for i in range(self.max_workers):
pool.spawn(self.work, i)
self.q.join()
end = time.time()
print("Finished crawling, takes %s seconds." % str(end - start))
print("Have fun hacking!")
class Interpreter(object):
'''
The class repsonsible for keeping track of the execution of the
statemachine.
'''
def __init__(self):
self.running = True
self.configuration = OrderedSet()
self.internalQueue = Queue()
self.externalQueue = Queue()
self.statesToInvoke = OrderedSet()
self.historyValue = {}
self.dm = None
self.invokeId = None
self.parentId = None
self.logger = None
def interpret(self, document, invokeId=None):
'''Initializes the interpreter given an SCXMLDocument instance'''
self.doc = document
self.invokeId = invokeId
transition = Transition(document.rootState)
transition.target = document.rootState.initial
transition.exe = document.rootState.initial.exe
self.executeTransitionContent([transition])
self.enterStates([transition])
def mainEventLoop(self):
while self.running:
enabledTransitions = None
stable = False
# now take any newly enabled null transitions and any transitions triggered by internal events
while self.running and not stable:
enabledTransitions = self.selectEventlessTransitions()
if not enabledTransitions:
if self.internalQueue.empty():
stable = True
else:
internalEvent = self.internalQueue.get() # this call returns immediately if no event is available
self.logger.info("internal event found: %s", internalEvent.name)
self.dm["__event"] = internalEvent
enabledTransitions = self.selectTransitions(internalEvent)
if enabledTransitions:
self.microstep(enabledTransitions)
# eventlet.greenthread.sleep()
eventlet.greenthread.sleep()
for state in self.statesToInvoke:
for inv in state.invoke:
inv.invoke(inv)
self.statesToInvoke.clear()
if not self.internalQueue.empty():
continue
externalEvent = self.externalQueue.get() # this call blocks until an event is available
if externalEvent.name == "cancel.invoke.%s" % self.dm.sessionid:
continue
self.logger.info("external event found: %s", externalEvent.name)
self.dm["__event"] = externalEvent
for state in self.configuration:
for inv in state.invoke:
if inv.invokeid == externalEvent.invokeid: # event is the result of an <invoke> in this state
self.applyFinalize(inv, externalEvent)
if inv.autoforward:
inv.send(externalEvent)
enabledTransitions = self.selectTransitions(externalEvent)
if enabledTransitions:
self.microstep(enabledTransitions)
# if we get here, we have reached a top-level final state or some external entity has set running to False
self.exitInterpreter()
def exitInterpreter(self):
statesToExit = sorted(self.configuration, key=exitOrder)
for s in statesToExit:
for content in s.onexit:
self.executeContent(content)
for inv in s.invoke:
self.cancelInvoke(inv)
self.configuration.delete(s)
if isFinalState(s) and isScxmlState(s.parent):
if self.invokeId and self.parentId and self.parentId in self.dm.sessions:
self.send(["done", "invoke", self.invokeId], s.donedata(), self.invokeId, self.dm.sessions[self.parentId].interpreter.externalQueue)
self.logger.info("Exiting interpreter")
dispatcher.send("signal_exit", self, final=s.id)
return
dispatcher.send("signal_exit", self, final=None)
def selectEventlessTransitions(self):
enabledTransitions = OrderedSet()
atomicStates = filter(isAtomicState, self.configuration)
atomicStates = sorted(atomicStates, key=documentOrder)
for state in atomicStates:
done = False
for s in [state] + getProperAncestors(state, None):
if done: break
for t in s.transition:
if not t.event and self.conditionMatch(t):
enabledTransitions.add(t)
done = True
break
filteredTransitions = self.filterPreempted(enabledTransitions)
return filteredTransitions
def selectTransitions(self, event):
enabledTransitions = OrderedSet()
atomicStates = filter(isAtomicState, self.configuration)
atomicStates = sorted(atomicStates, key=documentOrder)
for state in atomicStates:
done = False
for s in [state] + getProperAncestors(state, None):
if done: break
for t in s.transition:
if t.event and nameMatch(t.event, event.name.split(".")) and self.conditionMatch(t):
enabledTransitions.add(t)
done = True
break
filteredTransitions = self.filterPreempted(enabledTransitions)
return filteredTransitions
def preemptsTransition(self, t, t2):
if self.isType1(t): return False
elif self.isType2(t) and self.isType3(t2): return True
elif self.isType3(t): return True
return False
def getCommonParallel(self, states):
ancestors = set(getProperAncestors(states[0], None))
for s in states[1:]:
ancestors = ancestors.intersection(getProperAncestors(s, None))
if ancestors:
return sorted(ancestors, key=exitOrder)[0]
def isType1(self, t):
return not t.target
def isType2(self, t):
source = t.source if t.type == "internal" else t.source.parent
p = self.getCommonParallel([source] + self.getTargetStates(t.target))
return not isScxmlState(p)
def isType3(self, t):
return not self.isType2(t) and not self.isType1(t)
def filterPreempted(self, enabledTransitions):
filteredTransitions = []
for t in enabledTransitions:
# does any t2 in filteredTransitions preempt t? if not, add t to filteredTransitions
if not any(map(lambda t2: self.preemptsTransition(t2, t), filteredTransitions)):
filteredTransitions.append(t)
return OrderedSet(filteredTransitions)
def microstep(self, enabledTransitions):
self.exitStates(enabledTransitions)
self.executeTransitionContent(enabledTransitions)
self.enterStates(enabledTransitions)
self.logger.info("new config: {" + ", ".join([s.id for s in self.configuration if s.id != "__main__"]) + "}")
def exitStates(self, enabledTransitions):
statesToExit = OrderedSet()
for t in enabledTransitions:
if t.target:
tstates = self.getTargetStates(t.target)
if t.type == "internal" and isCompoundState(t.source) and all(map(lambda s: isDescendant(s,t.source), tstates)):
ancestor = t.source
else:
ancestor = self.findLCA([t.source] + tstates)
for s in self.configuration:
if isDescendant(s,ancestor):
statesToExit.add(s)
for s in statesToExit:
self.statesToInvoke.delete(s)
statesToExit.sort(key=exitOrder)
for s in statesToExit:
for h in s.history:
if h.type == "deep":
f = lambda s0: isAtomicState(s0) and isDescendant(s0,s)
else:
f = lambda s0: s0.parent == s
self.historyValue[h.id] = filter(f,self.configuration) #+ s.parent
for s in statesToExit:
for content in s.onexit:
self.executeContent(content)
for inv in s.invoke:
self.cancelInvoke(inv)
self.configuration.delete(s)
def cancelInvoke(self, inv):
inv.cancel()
def executeTransitionContent(self, enabledTransitions):
for t in enabledTransitions:
self.executeContent(t)
def enterStates(self, enabledTransitions):
statesToEnter = OrderedSet()
statesForDefaultEntry = OrderedSet()
for t in enabledTransitions:
if t.target:
tstates = self.getTargetStates(t.target)
if t.type == "internal" and isCompoundState(t.source) and all(map(lambda s: isDescendant(s,t.source), tstates)):
ancestor = t.source
else:
ancestor = self.findLCA([t.source] + tstates)
for s in tstates:
self.addStatesToEnter(s,statesToEnter,statesForDefaultEntry)
for s in tstates:
for anc in getProperAncestors(s,ancestor):
statesToEnter.add(anc)
if isParallelState(anc):
for child in getChildStates(anc):
if not any(map(lambda s: isDescendant(s,child), statesToEnter)):
self.addStatesToEnter(child, statesToEnter,statesForDefaultEntry)
statesToEnter.sort(key=enterOrder)
for s in statesToEnter:
self.statesToInvoke.add(s)
self.configuration.add(s)
if self.doc.binding == "late" and s.isFirstEntry:
s.initDatamodel()
s.isFirstEntry = False
for content in s.onentry:
self.executeContent(content)
if s in statesForDefaultEntry:
self.executeContent(s.initial)
if isFinalState(s):
parent = s.parent
grandparent = parent.parent
self.internalQueue.put(Event(["done", "state", parent.id], s.donedata()))
if isParallelState(grandparent):
if all(map(self.isInFinalState, getChildStates(grandparent))):
self.internalQueue.put(Event(["done", "state", grandparent.id]))
for s in self.configuration:
if isFinalState(s) and isScxmlState(s.parent):
self.running = False;
def addStatesToEnter(self, state,statesToEnter,statesForDefaultEntry):
if isHistoryState(state):
if state.id in self.historyValue:
for s in self.historyValue[state.id]:
self.addStatesToEnter(s, statesToEnter, statesForDefaultEntry)
for anc in getProperAncestors(s,state):
statesToEnter.add(anc)
else:
for t in state.transition:
for s in self.getTargetStates(t.target):
self.addStatesToEnter(s, statesToEnter, statesForDefaultEntry)
else:
statesToEnter.add(state)
if isCompoundState(state):
statesForDefaultEntry.add(state)
for s in self.getTargetStates(state.initial):
self.addStatesToEnter(s, statesToEnter, statesForDefaultEntry)
elif isParallelState(state):
for s in getChildStates(state):
self.addStatesToEnter(s,statesToEnter,statesForDefaultEntry)
def isInFinalState(self, s):
if isCompoundState(s):
return any(map(lambda s: isFinalState(s) and s in self.configuration, getChildStates(s)))
elif isParallelState(s):
return all(map(self.isInFinalState, getChildStates(s)))
else:
return False
def findLCA(self, stateList):
for anc in filter(isCompoundState, getProperAncestors(stateList[0], None)):
# for anc in getProperAncestors(stateList[0], None):
if all(map(lambda(s): isDescendant(s,anc), stateList[1:])):
return anc
class ECWriter(object):
"""
Writes an EC chunk
"""
def __init__(self, chunk, conn):
self._chunk = chunk
self._conn = conn
self.failed = False
self.bytes_transferred = 0
self.checksum = hashlib.md5()
@property
def chunk(self):
return self._chunk
@property
def conn(self):
return self._conn
@classmethod
def connect(cls, chunk, sysmeta):
raw_url = chunk["url"]
parsed = urlparse(raw_url)
chunk_path = parsed.path.split('/')[-1]
h = {}
h["transfer-encoding"] = "chunked"
h[chunk_headers["content_id"]] = sysmeta['id']
h[chunk_headers["content_path"]] = sysmeta['content_path']
h[chunk_headers["content_chunkmethod"]] = sysmeta['chunk_method']
h[chunk_headers["container_id"]] = sysmeta['container_id']
h[chunk_headers["chunk_pos"]] = chunk["pos"]
h[chunk_headers["chunk_id"]] = chunk_path
h[chunk_headers["content_policy"]] = sysmeta['policy']
h[chunk_headers["content_version"]] = sysmeta['version']
# in the trailer
# metachunk_size & metachunk_hash
h["Trailer"] = (chunk_headers["metachunk_size"],
chunk_headers["metachunk_hash"])
with ConnectionTimeout(io.CONNECTION_TIMEOUT):
conn = io.http_connect(parsed.netloc, 'PUT', parsed.path, h)
conn.chunk = chunk
return cls(chunk, conn)
def start(self, pool):
# we use eventlet Queue to pass data to the send coroutine
self.queue = Queue(io.PUT_QUEUE_DEPTH)
# spawn the send coroutine
pool.spawn(self._send)
def _send(self):
# this is the send coroutine loop
while True:
# fetch input data from the queue
d = self.queue.get()
# use HTTP transfer encoding chunked
# to write data to RAWX
if not self.failed:
# format the chunk
to_send = "%x\r\n%s\r\n" % (len(d), d)
try:
with ChunkWriteTimeout(io.CHUNK_TIMEOUT):
self.conn.send(to_send)
self.bytes_transferred += len(d)
except (Exception, ChunkWriteTimeout) as e:
self.failed = True
msg = str(e)
logger.warn("Failed to write to %s (%s)", self.chunk, msg)
self.chunk['error'] = msg
self.queue.task_done()
def wait(self):
# wait until all data in the queue
# has been processed by the send coroutine
if self.queue.unfinished_tasks:
self.queue.join()
def send(self, data):
# do not send empty data because
# this will end the chunked body
if not data:
return
# put the data to send into the queue
# it will be processed by the send coroutine
self.queue.put(data)
def finish(self, metachunk_size, metachunk_hash):
parts = [
'0\r\n',
'%s: %s\r\n' % (chunk_headers['metachunk_size'], metachunk_size),
'%s: %s\r\n' % (chunk_headers['metachunk_hash'], metachunk_hash),
'\r\n'
]
to_send = "".join(parts)
self.conn.send(to_send)
def getresponse(self):
# read the HTTP response from the connection
with Timeout(io.CHUNK_TIMEOUT):
self.resp = self.conn.getresponse()
return self.resp
