def test_init_cursor(self):
"""Test the init_cursor method
Cases:
1. no last checkpoint, no collection dump
2. no last checkpoint, collection dump ok and stuff to dump
3. no last checkpoint, nothing to dump, stuff in oplog
4. no last checkpoint, nothing to dump, nothing in oplog
5. no last checkpoint, no collection dump, stuff in oplog
6. last checkpoint exists
7. last checkpoint is behind
"""
# N.B. these sub-cases build off of each other and cannot be re-ordered
# without side-effects
# No last checkpoint, no collection dump, nothing in oplog
# "change oplog collection" to put nothing in oplog
self.opman1.oplog = self.shard1_conn["test"]["emptycollection"]
self.opman2.oplog = self.shard2_conn["test"]["emptycollection"]
self.opman1.collection_dump = False
self.opman2.collection_dump = False
self.assertTrue(
all(doc['op'] == 'n' for doc in self.opman1.init_cursor()[0]))
self.assertEqual(self.opman1.checkpoint, None)
self.assertTrue(
all(doc['op'] == 'n' for doc in self.opman2.init_cursor()[0]))
self.assertEqual(self.opman2.checkpoint, None)
# No last checkpoint, empty collections, nothing in oplog
self.opman1.collection_dump = self.opman2.collection_dump = True
cursor, cursor_len = self.opman1.init_cursor()
self.assertEqual(cursor, None)
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman1.checkpoint, None)
cursor, cursor_len = self.opman2.init_cursor()
self.assertEqual(cursor, None)
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman2.checkpoint, None)
# No last checkpoint, empty collections, something in oplog
self.opman1.oplog = self.shard1_conn["local"]["oplog.rs"]
self.opman2.oplog = self.shard2_conn["local"]["oplog.rs"]
oplog_startup_ts = self.opman2.get_last_oplog_timestamp()
collection = self.mongos_conn["test"]["mcsharded"]
collection.insert({"i": 1})
collection.remove({"i": 1})
time.sleep(3)
last_ts1 = self.opman1.get_last_oplog_timestamp()
cursor, cursor_len = self.opman1.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman1.checkpoint, last_ts1)
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)], last_ts1)
# init_cursor should point to startup message in shard2 oplog
cursor, cursor_len = self.opman2.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman2.checkpoint, oplog_startup_ts)
# No last checkpoint, no collection dump, stuff in oplog
progress = LockingDict()
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.collection_dump = self.opman2.collection_dump = False
collection.insert({"i": 1200})
last_ts2 = self.opman2.get_last_oplog_timestamp()
self.opman1.init_cursor()
self.assertEqual(self.opman1.checkpoint, last_ts1)
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)], last_ts1)
cursor, cursor_len = self.opman2.init_cursor()
for i in range(cursor_len - 1):
next(cursor)
self.assertEqual(next(cursor)["o"]["i"], 1200)
self.assertEqual(self.opman2.checkpoint, last_ts2)
with self.opman2.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman2.oplog)], last_ts2)
# Last checkpoint exists
progress = LockingDict()
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
for i in range(1000):
collection.insert({"i": i + 500})
entry1 = list(self.shard1_conn["local"]["oplog.rs"].find(skip=200,
limit=2))
entry2 = list(self.shard2_conn["local"]["oplog.rs"].find(skip=200,
limit=2))
progress.get_dict()[str(self.opman1.oplog)] = entry1[0]["ts"]
progress.get_dict()[str(self.opman2.oplog)] = entry2[0]["ts"]
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.checkpoint = self.opman2.checkpoint = None
cursor1, cursor_len1 = self.opman1.init_cursor()
cursor2, cursor_len2 = self.opman2.init_cursor()
self.assertEqual(entry1[1]["ts"], next(cursor1)["ts"])
self.assertEqual(entry2[1]["ts"], next(cursor2)["ts"])
self.assertEqual(self.opman1.checkpoint, entry1[0]["ts"])
self.assertEqual(self.opman2.checkpoint, entry2[0]["ts"])
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)],
entry1[0]["ts"])
with self.opman2.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman2.oplog)],
entry2[0]["ts"])
# Last checkpoint is behind
progress = LockingDict()
progress.get_dict()[str(self.opman1.oplog)] = bson.Timestamp(1, 0)
progress.get_dict()[str(self.opman2.oplog)] = bson.Timestamp(1, 0)
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.checkpoint = self.opman2.checkpoint = None
cursor, cursor_len = self.opman1.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(cursor, None)
self.assertIsNotNone(self.opman1.checkpoint)
cursor, cursor_len = self.opman2.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(cursor, None)
self.assertIsNotNone(self.opman2.checkpoint)
def test_init_cursor(self):
"""Test the init_cursor method
Cases:
1. no last checkpoint, no collection dump
2. no last checkpoint, collection dump ok and stuff to dump
3. no last checkpoint, nothing to dump, stuff in oplog
4. no last checkpoint, nothing to dump, nothing in oplog
5. no last checkpoint, no collection dump, stuff in oplog
6. last checkpoint exists
7. last checkpoint is behind
"""
# N.B. these sub-cases build off of each other and cannot be re-ordered
# without side-effects
# No last checkpoint, no collection dump, nothing in oplog
# "change oplog collection" to put nothing in oplog
self.opman.oplog = self.primary_conn["test"]["emptycollection"]
self.opman.collection_dump = False
self.assertTrue(
all(doc['op'] == 'n' for doc in self.opman.init_cursor()[0]))
self.assertEqual(self.opman.checkpoint, None)
# No last checkpoint, empty collections, nothing in oplog
self.opman.collection_dump = True
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(cursor, None)
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman.checkpoint, None)
# No last checkpoint, empty collections, something in oplog
self.opman.oplog = self.primary_conn['local']['oplog.rs']
collection = self.primary_conn["test"]["test"]
collection.insert_one({"i": 1})
collection.delete_one({"i": 1})
time.sleep(3)
last_ts = self.opman.get_last_oplog_timestamp()
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman.checkpoint, last_ts)
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)], last_ts)
# No last checkpoint, no collection dump, something in oplog
self.opman.oplog_progress = LockingDict()
self.opman.collection_dump = False
collection.insert_one({"i": 2})
last_ts = self.opman.get_last_oplog_timestamp()
cursor, cursor_len = self.opman.init_cursor()
for i in range(cursor_len - 1):
next(cursor)
self.assertEqual(next(cursor)['o']['i'], 2)
self.assertEqual(self.opman.checkpoint, last_ts)
# Last checkpoint exists
progress = LockingDict()
self.opman.oplog_progress = progress
for i in range(1000):
collection.insert_one({"i": i + 500})
entry = list(self.primary_conn["local"]["oplog.rs"].find(skip=200,
limit=-2))
progress.get_dict()[str(self.opman.oplog)] = entry[0]["ts"]
self.opman.oplog_progress = progress
self.opman.checkpoint = None
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(next(cursor)["ts"], entry[1]["ts"])
self.assertEqual(self.opman.checkpoint, entry[0]["ts"])
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)],
entry[0]["ts"])
# Last checkpoint is behind
progress = LockingDict()
progress.get_dict()[str(self.opman.oplog)] = bson.Timestamp(1, 0)
self.opman.oplog_progress = progress
self.opman.checkpoint = None
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(cursor, None)
self.assertIsNotNone(self.opman.checkpoint)
class Connector(threading.Thread):
"""Checks the cluster for shards to tail.
"""
def __init__(self, address, oplog_checkpoint, target_url, ns_set,
u_key, auth_key, doc_manager=None, auth_username=None,
collection_dump=True, batch_size=constants.DEFAULT_BATCH_SIZE,
fields=None, dest_mapping={},
auto_commit_interval=constants.DEFAULT_COMMIT_INTERVAL,
continue_on_error=False):
if target_url and not doc_manager:
raise errors.ConnectorError("Cannot create a Connector with a "
"target URL but no doc manager!")
def is_string(s):
try:
return isinstance(s, basestring)
except NameError:
return isinstance(s, str)
def load_doc_manager(path):
name, _ = os.path.splitext(os.path.basename(path))
try:
import importlib.machinery
loader = importlib.machinery.SourceFileLoader(name, path)
module = loader.load_module(name)
except ImportError:
module = imp.load_source(name, path)
return module
doc_manager_modules = None
if doc_manager is not None:
# backwards compatilibity: doc_manager may be a string
if is_string(doc_manager):
doc_manager_modules = [load_doc_manager(doc_manager)]
# doc_manager is a list
else:
doc_manager_modules = []
for dm in doc_manager:
doc_manager_modules.append(load_doc_manager(dm))
super(Connector, self).__init__()
#can_run is set to false when we join the thread
self.can_run = True
#The name of the file that stores the progress of the OplogThreads
self.oplog_checkpoint = oplog_checkpoint
#main address - either mongos for sharded setups or a primary otherwise
self.address = address
#The URLs of each target system, respectively
if is_string(target_url):
self.target_urls = [target_url]
elif target_url:
self.target_urls = list(target_url)
else:
self.target_urls = None
#The set of relevant namespaces to consider
self.ns_set = ns_set
#The dict of source namespace to destination namespace
self.dest_mapping = dest_mapping
#Whether the collection dump gracefully handles exceptions
self.continue_on_error = continue_on_error
#The key that is a unique document identifier for the target system.
#Not necessarily the mongo unique key.
self.u_key = u_key
#Password for authentication
self.auth_key = auth_key
#Username for authentication
self.auth_username = auth_username
#The set of OplogThreads created
self.shard_set = {}
#Boolean chooses whether to dump the entire collection if no timestamp
# is present in the config file
self.collection_dump = collection_dump
#Num entries to process before updating config file with current pos
self.batch_size = batch_size
#Dict of OplogThread/timestamp pairs to record progress
self.oplog_progress = LockingDict()
# List of fields to export
self.fields = fields
try:
docman_kwargs = {"unique_key": u_key,
"namespace_set": ns_set,
"auto_commit_interval": auto_commit_interval}
# No doc managers specified, using simulator
if doc_manager is None:
self.doc_managers = [simulator.DocManager(**docman_kwargs)]
else:
self.doc_managers = []
for i, d in enumerate(doc_manager_modules):
# self.target_urls may be shorter than
# self.doc_managers, or left as None
if self.target_urls and i < len(self.target_urls):
target_url = self.target_urls[i]
else:
target_url = None
if target_url:
self.doc_managers.append(
d.DocManager(self.target_urls[i],
**docman_kwargs))
else:
self.doc_managers.append(
d.DocManager(**docman_kwargs))
# If more target URLs were given than doc managers, may need
# to create additional doc managers
for url in self.target_urls[i + 1:]:
self.doc_managers.append(
doc_manager_modules[-1].DocManager(url,
**docman_kwargs))
except errors.ConnectionFailed:
err_msg = "MongoConnector: Could not connect to target system"
logging.critical(err_msg)
self.can_run = False
return
if self.oplog_checkpoint is not None:
if not os.path.exists(self.oplog_checkpoint):
info_str = ("MongoConnector: Can't find %s, "
"attempting to create an empty progress log" %
self.oplog_checkpoint)
logging.info(info_str)
try:
# Create oplog progress file
open(self.oplog_checkpoint, "w").close()
except IOError as e:
logging.critical("MongoConnector: Could not "
"create a progress log: %s" %
str(e))
sys.exit(2)
else:
if (not os.access(self.oplog_checkpoint, os.W_OK)
and not os.access(self.oplog_checkpoint, os.R_OK)):
logging.critical("Invalid permissions on %s! Exiting" %
(self.oplog_checkpoint))
sys.exit(2)
def join(self):
""" Joins thread, stops it from running
"""
self.can_run = False
for dm in self.doc_managers:
dm.stop()
threading.Thread.join(self)
def write_oplog_progress(self):
""" Writes oplog progress to file provided by user
"""
if self.oplog_checkpoint is None:
return None
# write to temp file
backup_file = self.oplog_checkpoint + '.backup'
os.rename(self.oplog_checkpoint, backup_file)
# for each of the threads write to file
with open(self.oplog_checkpoint, 'w') as dest:
with self.oplog_progress as oplog_prog:
oplog_dict = oplog_prog.get_dict()
for oplog, time_stamp in oplog_dict.items():
oplog_str = str(oplog)
timestamp = util.bson_ts_to_long(time_stamp)
json_str = json.dumps([oplog_str, timestamp])
try:
dest.write(json_str)
except IOError:
# Basically wipe the file, copy from backup
dest.truncate()
with open(backup_file, 'r') as backup:
shutil.copyfile(backup, dest)
break
os.remove(self.oplog_checkpoint + '.backup')
def read_oplog_progress(self):
"""Reads oplog progress from file provided by user.
This method is only called once before any threads are spanwed.
"""
if self.oplog_checkpoint is None:
return None
# Check for empty file
try:
if os.stat(self.oplog_checkpoint).st_size == 0:
logging.info("MongoConnector: Empty oplog progress file.")
return None
except OSError:
return None
source = open(self.oplog_checkpoint, 'r')
try:
data = json.load(source)
except ValueError: # empty file
reason = "It may be empty or corrupt."
logging.info("MongoConnector: Can't read oplog progress file. %s" %
(reason))
source.close()
return None
source.close()
count = 0
oplog_dict = self.oplog_progress.get_dict()
for count in range(0, len(data), 2):
oplog_str = data[count]
time_stamp = data[count + 1]
oplog_dict[oplog_str] = util.long_to_bson_ts(time_stamp)
#stored as bson_ts
def run(self):
"""Discovers the mongo cluster and creates a thread for each primary.
"""
main_conn = MongoClient(self.address)
if self.auth_key is not None:
main_conn['admin'].authenticate(self.auth_username, self.auth_key)
self.read_oplog_progress()
conn_type = None
try:
main_conn.admin.command("isdbgrid")
except pymongo.errors.OperationFailure:
conn_type = "REPLSET"
if conn_type == "REPLSET":
# Make sure we are connected to a replica set
is_master = main_conn.admin.command("isMaster")
if not "setName" in is_master:
logging.error(
'No replica set at "%s"! A replica set is required '
'to run mongo-connector. Shutting down...' % self.address
)
return
# Establish a connection to the replica set as a whole
main_conn.disconnect()
main_conn = MongoClient(self.address,
replicaSet=is_master['setName'])
if self.auth_key is not None:
main_conn.admin.authenticate(self.auth_username, self.auth_key)
#non sharded configuration
oplog_coll = main_conn['local']['oplog.rs']
oplog = OplogThread(
primary_conn=main_conn,
main_address=self.address,
oplog_coll=oplog_coll,
is_sharded=False,
doc_manager=self.doc_managers,
oplog_progress_dict=self.oplog_progress,
namespace_set=self.ns_set,
auth_key=self.auth_key,
auth_username=self.auth_username,
repl_set=is_master['setName'],
collection_dump=self.collection_dump,
batch_size=self.batch_size,
fields=self.fields,
dest_mapping=self.dest_mapping,
continue_on_error=self.continue_on_error
)
self.shard_set[0] = oplog
logging.info('MongoConnector: Starting connection thread %s' %
main_conn)
oplog.start()
while self.can_run:
if not self.shard_set[0].running:
logging.error("MongoConnector: OplogThread"
" %s unexpectedly stopped! Shutting down" %
(str(self.shard_set[0])))
self.oplog_thread_join()
for dm in self.doc_managers:
dm.stop()
return
self.write_oplog_progress()
time.sleep(1)
else: # sharded cluster
while self.can_run is True:
for shard_doc in main_conn['config']['shards'].find():
shard_id = shard_doc['_id']
if shard_id in self.shard_set:
if not self.shard_set[shard_id].running:
logging.error("MongoConnector: OplogThread "
"%s unexpectedly stopped! Shutting "
"down" %
(str(self.shard_set[shard_id])))
self.oplog_thread_join()
for dm in self.doc_managers:
dm.stop()
return
self.write_oplog_progress()
time.sleep(1)
continue
try:
repl_set, hosts = shard_doc['host'].split('/')
except ValueError:
cause = "The system only uses replica sets!"
logging.error("MongoConnector: %s", cause)
self.oplog_thread_join()
for dm in self.doc_managers:
dm.stop()
return
shard_conn = MongoClient(hosts, replicaSet=repl_set)
oplog_coll = shard_conn['local']['oplog.rs']
oplog = OplogThread(
primary_conn=shard_conn,
main_address=self.address,
oplog_coll=oplog_coll,
is_sharded=True,
doc_manager=self.doc_managers,
oplog_progress_dict=self.oplog_progress,
namespace_set=self.ns_set,
auth_key=self.auth_key,
auth_username=self.auth_username,
collection_dump=self.collection_dump,
batch_size=self.batch_size,
fields=self.fields,
dest_mapping=self.dest_mapping,
continue_on_error=self.continue_on_error
)
self.shard_set[shard_id] = oplog
msg = "Starting connection thread"
logging.info("MongoConnector: %s %s" % (msg, shard_conn))
oplog.start()
self.oplog_thread_join()
self.write_oplog_progress()
def oplog_thread_join(self):
"""Stops all the OplogThreads
"""
logging.info('MongoConnector: Stopping all OplogThreads')
for thread in self.shard_set.values():
thread.join()
def test_init_cursor(self):
"""Test the init_cursor method
Cases:
1. no last checkpoint, no collection dump
2. no last checkpoint, collection dump ok and stuff to dump
3. no last checkpoint, nothing to dump, stuff in oplog
4. no last checkpoint, nothing to dump, nothing in oplog
5. last checkpoint exists
"""
# N.B. these sub-cases build off of each other and cannot be re-ordered
# without side-effects
# No last checkpoint, no collection dump, nothing in oplog
# "change oplog collection" to put nothing in oplog
self.opman1.oplog = self.shard1_conn["test"]["emptycollection"]
self.opman2.oplog = self.shard2_conn["test"]["emptycollection"]
self.opman1.collection_dump = False
self.opman2.collection_dump = False
self.assertEqual(self.opman1.init_cursor(), None)
self.assertEqual(self.opman1.checkpoint, None)
self.assertEqual(self.opman2.init_cursor(), None)
self.assertEqual(self.opman2.checkpoint, None)
# No last checkpoint, empty collections, nothing in oplog
self.opman1.collection_dump = True
self.opman2.collection_dump = True
self.assertEqual(self.opman1.init_cursor(), None)
self.assertEqual(self.opman1.checkpoint, None)
self.assertEqual(self.opman2.init_cursor(), None)
self.assertEqual(self.opman2.checkpoint, None)
# No last checkpoint, empty collections, something in oplog
self.opman1.oplog = self.shard1_conn["local"]["oplog.rs"]
self.opman2.oplog = self.shard2_conn["local"]["oplog.rs"]
oplog_startup_ts = self.opman2.get_last_oplog_timestamp()
collection = self.mongos_conn["test"]["mcsharded"]
collection.insert({"i": 1})
collection.remove({"i": 1})
time.sleep(3)
last_ts1 = self.opman1.get_last_oplog_timestamp()
self.assertEqual(next(self.opman1.init_cursor())["ts"], last_ts1)
self.assertEqual(self.opman1.checkpoint, last_ts1)
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)], last_ts1)
# init_cursor should point to startup message in shard2 oplog
cursor = self.opman2.init_cursor()
self.assertEqual(next(cursor)["ts"], oplog_startup_ts)
self.assertEqual(self.opman2.checkpoint, oplog_startup_ts)
# No last checkpoint, non-empty collections, stuff in oplog
progress = LockingDict()
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
collection.insert({"i": 1200})
last_ts2 = self.opman2.get_last_oplog_timestamp()
self.assertEqual(next(self.opman1.init_cursor())["ts"], last_ts1)
self.assertEqual(self.opman1.checkpoint, last_ts1)
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)], last_ts1)
self.assertEqual(next(self.opman2.init_cursor())["ts"], last_ts2)
self.assertEqual(self.opman2.checkpoint, last_ts2)
with self.opman2.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman2.oplog)], last_ts2)
# Last checkpoint exists
progress = LockingDict()
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
for i in range(1000):
collection.insert({"i": i + 500})
entry1 = list(
self.shard1_conn["local"]["oplog.rs"].find(skip=200, limit=2))
entry2 = list(
self.shard2_conn["local"]["oplog.rs"].find(skip=200, limit=2))
progress.get_dict()[str(self.opman1.oplog)] = entry1[0]["ts"]
progress.get_dict()[str(self.opman2.oplog)] = entry2[0]["ts"]
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.checkpoint = self.opman2.checkpoint = None
cursor1 = self.opman1.init_cursor()
cursor2 = self.opman2.init_cursor()
self.assertEqual(entry1[1]["ts"], next(cursor1)["ts"])
self.assertEqual(entry2[1]["ts"], next(cursor2)["ts"])
self.assertEqual(self.opman1.checkpoint, entry1[0]["ts"])
self.assertEqual(self.opman2.checkpoint, entry2[0]["ts"])
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)],
entry1[0]["ts"])
with self.opman2.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman2.oplog)],
entry2[0]["ts"])
def test_init_cursor(self):
"""Test the init_cursor method
Cases:
1. no last checkpoint, no collection dump
2. no last checkpoint, collection dump ok and stuff to dump
3. no last checkpoint, nothing to dump, stuff in oplog
4. no last checkpoint, nothing to dump, nothing in oplog
5. no last checkpoint, no collection dump, stuff in oplog
6. last checkpoint exists
7. last checkpoint is behind
"""
# N.B. these sub-cases build off of each other and cannot be re-ordered
# without side-effects
# No last checkpoint, no collection dump, nothing in oplog
# "change oplog collection" to put nothing in oplog
self.opman.oplog = self.primary_conn["test"]["emptycollection"]
self.opman.collection_dump = False
self.assertTrue(all(doc["op"] == "n" for doc in self.opman.init_cursor()[0]))
self.assertEqual(self.opman.checkpoint, None)
# No last checkpoint, empty collections, nothing in oplog
self.opman.collection_dump = True
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(cursor, None)
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman.checkpoint, None)
# No last checkpoint, empty collections, something in oplog
self.opman.oplog = self.primary_conn["local"]["oplog.rs"]
collection = self.primary_conn["test"]["test"]
collection.insert({"i": 1})
collection.remove({"i": 1})
time.sleep(3)
last_ts = self.opman.get_last_oplog_timestamp()
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman.checkpoint, last_ts)
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)], last_ts)
# No last checkpoint, no collection dump, something in oplog
self.opman.oplog_progress = LockingDict()
self.opman.collection_dump = False
collection.insert({"i": 2})
last_ts = self.opman.get_last_oplog_timestamp()
cursor, cursor_len = self.opman.init_cursor()
for i in range(cursor_len - 1):
next(cursor)
self.assertEqual(next(cursor)["o"]["i"], 2)
self.assertEqual(self.opman.checkpoint, last_ts)
# Last checkpoint exists
progress = LockingDict()
self.opman.oplog_progress = progress
for i in range(1000):
collection.insert({"i": i + 500})
entry = list(self.primary_conn["local"]["oplog.rs"].find(skip=200, limit=2))
progress.get_dict()[str(self.opman.oplog)] = entry[0]["ts"]
self.opman.oplog_progress = progress
self.opman.checkpoint = None
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(next(cursor)["ts"], entry[1]["ts"])
self.assertEqual(self.opman.checkpoint, entry[0]["ts"])
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)], entry[0]["ts"])
# Last checkpoint is behind
progress = LockingDict()
progress.get_dict()[str(self.opman.oplog)] = bson.Timestamp(1, 0)
self.opman.oplog_progress = progress
self.opman.checkpoint = None
cursor, cursor_len = self.opman.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(cursor, None)
self.assertIsNotNone(self.opman.checkpoint)
class Connector(threading.Thread):
"""Checks the cluster for shards to tail.
"""
def __init__(self,
address,
oplog_checkpoint,
target_url,
ns_set,
u_key,
auth_key,
doc_manager=None,
auth_username=None,
collection_dump=True,
batch_size=constants.DEFAULT_BATCH_SIZE,
fields=None,
dest_mapping={},
auto_commit_interval=constants.DEFAULT_COMMIT_INTERVAL,
continue_on_error=False):
if target_url and not doc_manager:
raise errors.ConnectorError("Cannot create a Connector with a "
"target URL but no doc manager!")
def is_string(s):
try:
return isinstance(s, basestring)
except NameError:
return isinstance(s, str)
def load_doc_manager(path):
name, _ = os.path.splitext(os.path.basename(path))
try:
import importlib.machinery
loader = importlib.machinery.SourceFileLoader(name, path)
module = loader.load_module(name)
except ImportError:
module = imp.load_source(name, path)
return module
doc_manager_modules = None
if doc_manager is not None:
# backwards compatilibity: doc_manager may be a string
if is_string(doc_manager):
doc_manager_modules = [load_doc_manager(doc_manager)]
# doc_manager is a list
else:
doc_manager_modules = []
for dm in doc_manager:
doc_manager_modules.append(load_doc_manager(dm))
super(Connector, self).__init__()
#can_run is set to false when we join the thread
self.can_run = True
#The name of the file that stores the progress of the OplogThreads
self.oplog_checkpoint = oplog_checkpoint
#main address - either mongos for sharded setups or a primary otherwise
self.address = address
#The URLs of each target system, respectively
if is_string(target_url):
self.target_urls = [target_url]
elif target_url:
self.target_urls = list(target_url)
else:
self.target_urls = None
#The set of relevant namespaces to consider
self.ns_set = ns_set
#The dict of source namespace to destination namespace
self.dest_mapping = dest_mapping
#Whether the collection dump gracefully handles exceptions
self.continue_on_error = continue_on_error
#The key that is a unique document identifier for the target system.
#Not necessarily the mongo unique key.
self.u_key = u_key
#Password for authentication
self.auth_key = auth_key
#Username for authentication
self.auth_username = auth_username
#The set of OplogThreads created
self.shard_set = {}
#Boolean chooses whether to dump the entire collection if no timestamp
# is present in the config file
self.collection_dump = collection_dump
#Num entries to process before updating config file with current pos
self.batch_size = batch_size
#Dict of OplogThread/timestamp pairs to record progress
self.oplog_progress = LockingDict()
# List of fields to export
self.fields = fields
try:
docman_kwargs = {
"unique_key": u_key,
"namespace_set": ns_set,
"auto_commit_interval": auto_commit_interval
}
# No doc managers specified, using simulator
if doc_manager is None:
self.doc_managers = [simulator.DocManager(**docman_kwargs)]
else:
self.doc_managers = []
for i, d in enumerate(doc_manager_modules):
# self.target_urls may be shorter than
# self.doc_managers, or left as None
if self.target_urls and i < len(self.target_urls):
target_url = self.target_urls[i]
else:
target_url = None
if target_url:
self.doc_managers.append(
d.DocManager(self.target_urls[i], **docman_kwargs))
else:
self.doc_managers.append(d.DocManager(**docman_kwargs))
# If more target URLs were given than doc managers, may need
# to create additional doc managers
for url in self.target_urls[i + 1:]:
self.doc_managers.append(
doc_manager_modules[-1].DocManager(
url, **docman_kwargs))
except errors.ConnectionFailed:
err_msg = "MongoConnector: Could not connect to target system"
logging.critical(err_msg)
self.can_run = False
return
if self.oplog_checkpoint is not None:
if not os.path.exists(self.oplog_checkpoint):
info_str = ("MongoConnector: Can't find %s, "
"attempting to create an empty progress log" %
self.oplog_checkpoint)
logging.info(info_str)
try:
# Create oplog progress file
open(self.oplog_checkpoint, "w").close()
except IOError as e:
logging.critical("MongoConnector: Could not "
"create a progress log: %s" % str(e))
sys.exit(2)
else:
if (not os.access(self.oplog_checkpoint, os.W_OK)
and not os.access(self.oplog_checkpoint, os.R_OK)):
logging.critical("Invalid permissions on %s! Exiting" %
(self.oplog_checkpoint))
sys.exit(2)
def join(self):
""" Joins thread, stops it from running
"""
self.can_run = False
for dm in self.doc_managers:
dm.stop()
threading.Thread.join(self)
def write_oplog_progress(self):
""" Writes oplog progress to file provided by user
"""
if self.oplog_checkpoint is None:
return None
# write to temp file
backup_file = self.oplog_checkpoint + '.backup'
os.rename(self.oplog_checkpoint, backup_file)
# for each of the threads write to file
with open(self.oplog_checkpoint, 'w') as dest:
with self.oplog_progress as oplog_prog:
oplog_dict = oplog_prog.get_dict()
for oplog, time_stamp in oplog_dict.items():
oplog_str = str(oplog)
timestamp = util.bson_ts_to_long(time_stamp)
json_str = json.dumps([oplog_str, timestamp])
try:
dest.write(json_str)
except IOError:
# Basically wipe the file, copy from backup
dest.truncate()
with open(backup_file, 'r') as backup:
shutil.copyfile(backup, dest)
break
os.remove(self.oplog_checkpoint + '.backup')
def read_oplog_progress(self):
"""Reads oplog progress from file provided by user.
This method is only called once before any threads are spanwed.
"""
if self.oplog_checkpoint is None:
return None
# Check for empty file
try:
if os.stat(self.oplog_checkpoint).st_size == 0:
logging.info("MongoConnector: Empty oplog progress file.")
return None
except OSError:
return None
source = open(self.oplog_checkpoint, 'r')
try:
data = json.load(source)
except ValueError: # empty file
reason = "It may be empty or corrupt."
logging.info("MongoConnector: Can't read oplog progress file. %s" %
(reason))
source.close()
return None
source.close()
count = 0
oplog_dict = self.oplog_progress.get_dict()
for count in range(0, len(data), 2):
oplog_str = data[count]
time_stamp = data[count + 1]
oplog_dict[oplog_str] = util.long_to_bson_ts(time_stamp)
#stored as bson_ts
def run(self):
"""Discovers the mongo cluster and creates a thread for each primary.
"""
main_conn = MongoClient(self.address)
if self.auth_key is not None:
main_conn['admin'].authenticate(self.auth_username, self.auth_key)
self.read_oplog_progress()
conn_type = None
try:
main_conn.admin.command("isdbgrid")
except pymongo.errors.OperationFailure:
conn_type = "REPLSET"
if conn_type == "REPLSET":
# Make sure we are connected to a replica set
is_master = main_conn.admin.command("isMaster")
if not "setName" in is_master:
logging.error(
'No replica set at "%s"! A replica set is required '
'to run mongo-connector. Shutting down...' % self.address)
return
# Establish a connection to the replica set as a whole
main_conn.disconnect()
main_conn = MongoClient(self.address,
replicaSet=is_master['setName'])
if self.auth_key is not None:
main_conn.admin.authenticate(self.auth_username, self.auth_key)
#non sharded configuration
oplog_coll = main_conn['local']['oplog.rs']
oplog = OplogThread(primary_conn=main_conn,
main_address=self.address,
oplog_coll=oplog_coll,
is_sharded=False,
doc_manager=self.doc_managers,
oplog_progress_dict=self.oplog_progress,
namespace_set=self.ns_set,
auth_key=self.auth_key,
auth_username=self.auth_username,
repl_set=is_master['setName'],
collection_dump=self.collection_dump,
batch_size=self.batch_size,
fields=self.fields,
dest_mapping=self.dest_mapping,
continue_on_error=self.continue_on_error)
self.shard_set[0] = oplog
logging.info('MongoConnector: Starting connection thread %s' %
main_conn)
oplog.start()
while self.can_run:
if not self.shard_set[0].running:
logging.error("MongoConnector: OplogThread"
" %s unexpectedly stopped! Shutting down" %
(str(self.shard_set[0])))
self.oplog_thread_join()
for dm in self.doc_managers:
dm.stop()
return
self.write_oplog_progress()
time.sleep(1)
else: # sharded cluster
while self.can_run is True:
for shard_doc in main_conn['config']['shards'].find():
shard_id = shard_doc['_id']
if shard_id in self.shard_set:
if not self.shard_set[shard_id].running:
logging.error("MongoConnector: OplogThread "
"%s unexpectedly stopped! Shutting "
"down" %
(str(self.shard_set[shard_id])))
self.oplog_thread_join()
for dm in self.doc_managers:
dm.stop()
return
self.write_oplog_progress()
time.sleep(1)
continue
try:
repl_set, hosts = shard_doc['host'].split('/')
except ValueError:
cause = "The system only uses replica sets!"
logging.error("MongoConnector: %s", cause)
self.oplog_thread_join()
for dm in self.doc_managers:
dm.stop()
return
shard_conn = MongoClient(hosts, replicaSet=repl_set)
oplog_coll = shard_conn['local']['oplog.rs']
oplog = OplogThread(
primary_conn=shard_conn,
main_address=self.address,
oplog_coll=oplog_coll,
is_sharded=True,
doc_manager=self.doc_managers,
oplog_progress_dict=self.oplog_progress,
namespace_set=self.ns_set,
auth_key=self.auth_key,
auth_username=self.auth_username,
collection_dump=self.collection_dump,
batch_size=self.batch_size,
fields=self.fields,
dest_mapping=self.dest_mapping,
continue_on_error=self.continue_on_error)
self.shard_set[shard_id] = oplog
msg = "Starting connection thread"
logging.info("MongoConnector: %s %s" % (msg, shard_conn))
oplog.start()
self.oplog_thread_join()
self.write_oplog_progress()
def oplog_thread_join(self):
"""Stops all the OplogThreads
"""
logging.info('MongoConnector: Stopping all OplogThreads')
for thread in self.shard_set.values():
thread.join()
def test_init_cursor(self):
"""Test the init_cursor method
Cases:
1. no last checkpoint, no collection dump
2. no last checkpoint, collection dump ok and stuff to dump
3. no last checkpoint, nothing to dump, stuff in oplog
4. no last checkpoint, nothing to dump, nothing in oplog
5. no last checkpoint, no collection dump, stuff in oplog
6. last checkpoint exists
7. last checkpoint is behind
"""
# N.B. these sub-cases build off of each other and cannot be re-ordered
# without side-effects
# No last checkpoint, no collection dump, nothing in oplog
# "change oplog collection" to put nothing in oplog
self.opman1.oplog = self.shard1_conn["test"]["emptycollection"]
self.opman2.oplog = self.shard2_conn["test"]["emptycollection"]
self.opman1.collection_dump = False
self.opman2.collection_dump = False
self.assertTrue(all(doc['op'] == 'n'
for doc in self.opman1.init_cursor()[0]))
self.assertEqual(self.opman1.checkpoint, None)
self.assertTrue(all(doc['op'] == 'n'
for doc in self.opman2.init_cursor()[0]))
self.assertEqual(self.opman2.checkpoint, None)
# No last checkpoint, empty collections, nothing in oplog
self.opman1.collection_dump = self.opman2.collection_dump = True
cursor, cursor_len = self.opman1.init_cursor()
self.assertEqual(cursor, None)
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman1.checkpoint, None)
cursor, cursor_len = self.opman2.init_cursor()
self.assertEqual(cursor, None)
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman2.checkpoint, None)
# No last checkpoint, empty collections, something in oplog
self.opman1.oplog = self.shard1_conn["local"]["oplog.rs"]
self.opman2.oplog = self.shard2_conn["local"]["oplog.rs"]
oplog_startup_ts = self.opman2.get_last_oplog_timestamp()
collection = self.mongos_conn["test"]["mcsharded"]
collection.insert_one({"i": 1})
collection.delete_one({"i": 1})
time.sleep(3)
last_ts1 = self.opman1.get_last_oplog_timestamp()
cursor, cursor_len = self.opman1.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman1.checkpoint, last_ts1)
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)], last_ts1)
# init_cursor should point to startup message in shard2 oplog
cursor, cursor_len = self.opman2.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(self.opman2.checkpoint, oplog_startup_ts)
# No last checkpoint, no collection dump, stuff in oplog
progress = LockingDict()
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.collection_dump = self.opman2.collection_dump = False
collection.insert_one({"i": 1200})
last_ts2 = self.opman2.get_last_oplog_timestamp()
self.opman1.init_cursor()
self.assertEqual(self.opman1.checkpoint, last_ts1)
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)], last_ts1)
cursor, cursor_len = self.opman2.init_cursor()
for i in range(cursor_len - 1):
next(cursor)
self.assertEqual(next(cursor)["o"]["i"], 1200)
self.assertEqual(self.opman2.checkpoint, last_ts2)
with self.opman2.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman2.oplog)], last_ts2)
# Last checkpoint exists
progress = LockingDict()
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
for i in range(1000):
collection.insert_one({"i": i + 500})
entry1 = list(
self.shard1_conn["local"]["oplog.rs"].find(skip=200, limit=-2))
entry2 = list(
self.shard2_conn["local"]["oplog.rs"].find(skip=200, limit=-2))
progress.get_dict()[str(self.opman1.oplog)] = entry1[0]["ts"]
progress.get_dict()[str(self.opman2.oplog)] = entry2[0]["ts"]
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.checkpoint = self.opman2.checkpoint = None
cursor1, cursor_len1 = self.opman1.init_cursor()
cursor2, cursor_len2 = self.opman2.init_cursor()
self.assertEqual(entry1[1]["ts"], next(cursor1)["ts"])
self.assertEqual(entry2[1]["ts"], next(cursor2)["ts"])
self.assertEqual(self.opman1.checkpoint, entry1[0]["ts"])
self.assertEqual(self.opman2.checkpoint, entry2[0]["ts"])
with self.opman1.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman1.oplog)],
entry1[0]["ts"])
with self.opman2.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman2.oplog)],
entry2[0]["ts"])
# Last checkpoint is behind
progress = LockingDict()
progress.get_dict()[str(self.opman1.oplog)] = bson.Timestamp(1, 0)
progress.get_dict()[str(self.opman2.oplog)] = bson.Timestamp(1, 0)
self.opman1.oplog_progress = self.opman2.oplog_progress = progress
self.opman1.checkpoint = self.opman2.checkpoint = None
cursor, cursor_len = self.opman1.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(cursor, None)
self.assertIsNotNone(self.opman1.checkpoint)
cursor, cursor_len = self.opman2.init_cursor()
self.assertEqual(cursor_len, 0)
self.assertEqual(cursor, None)
self.assertIsNotNone(self.opman2.checkpoint)
def test_init_cursor(self):
"""Test the init_cursor method
Cases:
1. no last checkpoint, no collection dump
2. no last checkpoint, collection dump ok and stuff to dump
3. no last checkpoint, nothing to dump, stuff in oplog
4. no last checkpoint, nothing to dump, nothing in oplog
5. last checkpoint exists
"""
# N.B. these sub-cases build off of each other and cannot be re-ordered
# without side-effects
# No last checkpoint, no collection dump, nothing in oplog
# "change oplog collection" to put nothing in oplog
self.opman.oplog = self.primary_conn["test"]["emptycollection"]
self.opman.collection_dump = False
self.assertEqual(self.opman.init_cursor(), None)
self.assertEqual(self.opman.checkpoint, None)
# No last checkpoint, empty collections, nothing in oplog
self.opman.collection_dump = True
self.assertEqual(self.opman.init_cursor(), None)
self.assertEqual(self.opman.checkpoint, None)
# No last checkpoint, empty collections, something in oplog
self.opman.oplog = self.primary_conn['local']['oplog.rs']
collection = self.primary_conn["test"]["test"]
collection.insert({"i": 1})
collection.remove({"i": 1})
time.sleep(3)
last_ts = self.opman.get_last_oplog_timestamp()
self.assertEqual(next(self.opman.init_cursor())["ts"], last_ts)
self.assertEqual(self.opman.checkpoint, last_ts)
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)], last_ts)
# No last checkpoint, non-empty collections, stuff in oplog
self.opman.oplog_progress = LockingDict()
self.assertEqual(next(self.opman.init_cursor())["ts"], last_ts)
self.assertEqual(self.opman.checkpoint, last_ts)
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)], last_ts)
# Last checkpoint exists
progress = LockingDict()
self.opman.oplog_progress = progress
for i in range(1000):
collection.insert({"i": i + 500})
entry = list(
self.primary_conn["local"]["oplog.rs"].find(skip=200, limit=2))
progress.get_dict()[str(self.opman.oplog)] = entry[0]["ts"]
self.opman.oplog_progress = progress
self.opman.checkpoint = None
cursor = self.opman.init_cursor()
self.assertEqual(entry[1]["ts"], next(cursor)["ts"])
self.assertEqual(self.opman.checkpoint, entry[0]["ts"])
with self.opman.oplog_progress as prog:
self.assertEqual(prog.get_dict()[str(self.opman.oplog)],
entry[0]["ts"])
class Connector(threading.Thread):
"""Checks the cluster for shards to tail.
"""
def __init__(self, address, oplog_checkpoint, target_url, ns_set,
u_key, auth_key, doc_manager=None, auth_username=None):
if doc_manager is not None:
doc_manager = imp.load_source('DocManager', doc_manager)
else:
from mongo_connector.doc_manager import DocManager
time.sleep(1)
super(Connector, self).__init__()
#can_run is set to false when we join the thread
self.can_run = True
#The name of the file that stores the progress of the OplogThreads
self.oplog_checkpoint = oplog_checkpoint
#main address - either mongos for sharded setups or a primary otherwise
self.address = address
#The URL of the target system
self.target_url = target_url
#The set of relevant namespaces to consider
self.ns_set = ns_set
#The key that is a unique document identifier for the target system.
#Not necessarily the mongo unique key.
self.u_key = u_key
#Password for authentication
self.auth_key = auth_key
#Username for authentication
self.auth_username = auth_username
#The set of OplogThreads created
self.shard_set = {}
#Dict of OplogThread/timestamp pairs to record progress
self.oplog_progress = LockingDict()
try:
if target_url is None:
if doc_manager is None: # imported using from... import
self.doc_manager = DocManager(unique_key=u_key)
else: # imported using load source
self.doc_manager = doc_manager.DocManager(unique_key=u_key)
else:
if doc_manager is None:
self.doc_manager = DocManager(self.target_url,
unique_key=u_key)
else:
self.doc_manager = doc_manager.DocManager(self.target_url,
unique_key=u_key)
except errors.ConnectionFailed:
err_msg = "MongoConnector: Could not connect to target system"
logging.critical(err_msg)
self.can_run = False
return
if self.oplog_checkpoint is not None:
if not os.path.exists(self.oplog_checkpoint):
info_str = ("MongoConnector: Can't find %s, "
"attempting to create an empty progress log" %
self.oplog_checkpoint)
logging.info(info_str)
try:
# Create oplog progress file
open(self.oplog_checkpoint, "w").close()
except IOError as e:
logging.critical("MongoConnector: Could not "
"create a progress log: %s" %
str(e))
sys.exit(1)
else:
if (not os.access(self.oplog_checkpoint, os.W_OK)
and not os.access(self.oplog_checkpoint, os.R_OK )):
logging.critical("Invalid permissions on %s! Exiting" %
(self.oplog_checkpoint))
sys.exit(1)
def join(self):
""" Joins thread, stops it from running
"""
self.can_run = False
self.doc_manager.stop()
threading.Thread.join(self)
def write_oplog_progress(self):
""" Writes oplog progress to file provided by user
"""
if self.oplog_checkpoint is None:
return None
# write to temp file
backup_file = self.oplog_checkpoint + '.backup'
os.rename(self.oplog_checkpoint, backup_file)
# for each of the threads write to file
with open(self.oplog_checkpoint, 'w') as dest:
with self.oplog_progress as oplog_prog:
oplog_dict = oplog_prog.get_dict()
for oplog, time_stamp in oplog_dict.items():
oplog_str = str(oplog)
timestamp = util.bson_ts_to_long(time_stamp)
json_str = json.dumps([oplog_str, timestamp])
try:
dest.write(json_str)
except IOError:
# Basically wipe the file, copy from backup
dest.truncate()
with open(backup_file, 'r') as backup:
shutil.copyfile(backup, dest)
break
os.remove(self.oplog_checkpoint + '.backup')
def read_oplog_progress(self):
"""Reads oplog progress from file provided by user.
This method is only called once before any threads are spanwed.
"""
if self.oplog_checkpoint is None:
return None
# Check for empty file
try:
if os.stat(self.oplog_checkpoint).st_size == 0:
logging.info("MongoConnector: Empty oplog progress file.")
return None
except OSError:
return None
source = open(self.oplog_checkpoint, 'r')
try:
data = json.load(source)
except ValueError: # empty file
reason = "It may be empty or corrupt."
logging.info("MongoConnector: Can't read oplog progress file. %s" %
(reason))
source.close()
return None
source.close()
count = 0
oplog_dict = self.oplog_progress.get_dict()
for count in range(0, len(data), 2):
oplog_str = data[count]
time_stamp = data[count + 1]
oplog_dict[oplog_str] = util.long_to_bson_ts(time_stamp)
#stored as bson_ts
def run(self):
"""Discovers the mongo cluster and creates a thread for each primary.
"""
main_conn = Connection(self.address)
if self.auth_key is not None:
main_conn['admin'].authenticate(self.auth_username, self.auth_key)
self.read_oplog_progress()
conn_type = None
try:
main_conn.admin.command("isdbgrid")
except pymongo.errors.OperationFailure:
conn_type = "REPLSET"
if conn_type == "REPLSET":
#non sharded configuration
oplog_coll = main_conn['local']['oplog.rs']
prim_admin = main_conn.admin
repl_set = prim_admin.command("replSetGetStatus")['set']
oplog = oplog_manager.OplogThread(main_conn,
(main_conn.host + ":" + str(main_conn.port)),
oplog_coll,
False, self.doc_manager,
self.oplog_progress,
self.ns_set, self.auth_key,
self.auth_username,
repl_set=repl_set)
self.shard_set[0] = oplog
logging.info('MongoConnector: Starting connection thread %s' %
main_conn)
oplog.start()
while self.can_run:
if not self.shard_set[0].running:
logging.error("MongoConnector: OplogThread"
" %s unexpectedly stopped! Shutting down" %
(str(self.shard_set[0])))
self.oplog_thread_join()
self.doc_manager.stop()
return
self.write_oplog_progress()
time.sleep(1)
else: # sharded cluster
while self.can_run is True:
for shard_doc in main_conn['config']['shards'].find():
shard_id = shard_doc['_id']
if shard_id in self.shard_set:
if not self.shard_set[shard_id].running:
logging.error("MongoConnector: OplogThread"
" %s unexpectedly stopped! Shutting down" %
(str(self.shard_set[shard_id])))
self.oplog_thread_join()
self.doc_manager.stop()
return
self.write_oplog_progress()
time.sleep(1)
continue
try:
repl_set, hosts = shard_doc['host'].split('/')
except ValueError:
cause = "The system only uses replica sets!"
logging.error("MongoConnector: %s", cause)
self.oplog_thread_join()
self.doc_manager.stop()
return
shard_conn = Connection(hosts, replicaset=repl_set)
oplog_coll = shard_conn['local']['oplog.rs']
oplog = oplog_manager.OplogThread(shard_conn, self.address,
oplog_coll, True,
self.doc_manager,
self.oplog_progress,
self.ns_set,
self.auth_key,
self.auth_username)
self.shard_set[shard_id] = oplog
msg = "Starting connection thread"
logging.info("MongoConnector: %s %s" % (msg, shard_conn))
oplog.start()
self.oplog_thread_join()
self.write_oplog_progress()
def oplog_thread_join(self):
"""Stops all the OplogThreads
"""
logging.info('MongoConnector: Stopping all OplogThreads')
for thread in self.shard_set.values():
thread.join()
