def dump_collection(self):
"""Dumps collection into the target system.
This method is called when we're initializing the cursor and have no
configs i.e. when we're starting for the first time.
"""
timestamp = retry_until_ok(self.get_last_oplog_timestamp)
if timestamp is None:
return None
long_ts = util.bson_ts_to_long(timestamp)
# Flag if this oplog thread was cancelled during the collection dump.
# Use a list to workaround python scoping.
dump_cancelled = [False]
def get_all_ns():
ns_set = []
gridfs_ns_set = []
db_list = self.namespace_config.get_included_databases()
if not db_list:
# Only use listDatabases when the configured databases are not
# explicit.
db_list = retry_until_ok(self.primary_client.database_names)
for database in db_list:
if database == "config" or database == "local":
continue
coll_list = retry_until_ok(
self.primary_client[database].collection_names
)
for coll in coll_list:
# ignore system collections
if coll.startswith("system."):
continue
# ignore gridfs chunks collections
if coll.endswith(".chunks"):
continue
if coll.endswith(".files"):
namespace = "%s.%s" % (database, coll)
namespace = namespace[: -len(".files")]
if self.namespace_config.gridfs_namespace(namespace):
gridfs_ns_set.append(namespace)
else:
namespace = "%s.%s" % (database, coll)
if self.namespace_config.map_namespace(namespace):
ns_set.append(namespace)
return ns_set, gridfs_ns_set
dump_set, gridfs_dump_set = get_all_ns()
LOG.debug("OplogThread: Dumping set of collections %s " % dump_set)
def docs_to_dump(from_coll):
last_id = None
attempts = 0
projection = self.namespace_config.projection(from_coll.full_name)
# Loop to handle possible AutoReconnect
while attempts < 60:
if last_id is None:
cursor = retry_until_ok(
from_coll.find,
projection=projection,
sort=[("_id", pymongo.ASCENDING)],
)
else:
cursor = retry_until_ok(
from_coll.find,
{"_id": {"$gt": last_id}},
projection=projection,
sort=[("_id", pymongo.ASCENDING)],
)
try:
for doc in cursor:
if not self.running:
# Thread was joined while performing the
# collection dump.
dump_cancelled[0] = True
raise StopIteration
last_id = doc["_id"]
yield doc
break
except (pymongo.errors.AutoReconnect, pymongo.errors.OperationFailure):
attempts += 1
time.sleep(1)
def upsert_each(dm):
num_failed = 0
for namespace in dump_set:
from_coll = self.get_collection(namespace)
mapped_ns = self.namespace_config.map_namespace(namespace)
total_docs = retry_until_ok(from_coll.count)
num = None
for num, doc in enumerate(docs_to_dump(from_coll)):
try:
dm.upsert(doc, mapped_ns, long_ts)
except Exception:
if self.continue_on_error:
LOG.exception("Could not upsert document: %r" % doc)
num_failed += 1
else:
raise
if num % 10000 == 0:
LOG.info(
"Upserted %d out of approximately %d docs "
"from collection '%s'",
num + 1,
total_docs,
namespace,
)
if num is not None:
LOG.info(
"Upserted %d out of approximately %d docs from "
"collection '%s'",
num + 1,
total_docs,
namespace,
)
if num_failed > 0:
LOG.error("Failed to upsert %d docs" % num_failed)
def upsert_all(dm):
try:
for namespace in dump_set:
from_coll = self.get_collection(namespace)
total_docs = retry_until_ok(from_coll.count)
mapped_ns = self.namespace_config.map_namespace(namespace)
LOG.info(
"Bulk upserting approximately %d docs from " "collection '%s'",
total_docs,
namespace,
)
dm.bulk_upsert(docs_to_dump(from_coll), mapped_ns, long_ts)
except Exception:
if self.continue_on_error:
LOG.exception(
"OplogThread: caught exception"
" during bulk upsert, re-upserting"
" documents serially"
)
upsert_each(dm)
else:
raise
def do_dump(dm, error_queue):
try:
LOG.debug(
"OplogThread: Using bulk upsert function for " "collection dump"
)
upsert_all(dm)
if gridfs_dump_set:
LOG.info(
"OplogThread: dumping GridFS collections: %s", gridfs_dump_set
)
# Dump GridFS files
for gridfs_ns in gridfs_dump_set:
mongo_coll = self.get_collection(gridfs_ns)
from_coll = self.get_collection(gridfs_ns + ".files")
dest_ns = self.namespace_config.map_namespace(gridfs_ns)
for doc in docs_to_dump(from_coll):
gridfile = GridFSFile(mongo_coll, doc)
dm.insert_file(gridfile, dest_ns, long_ts)
except Exception:
# Likely exceptions:
# pymongo.errors.OperationFailure,
# mongo_connector.errors.ConnectionFailed
# mongo_connector.errors.OperationFailed
error_queue.put(sys.exc_info())
# Extra threads (if any) that assist with collection dumps
dumping_threads = []
# Did the dump succeed for all target systems?
dump_success = True
# Holds any exceptions we can't recover from
errors = queue.Queue()
if len(self.doc_managers) == 1:
do_dump(self.doc_managers[0], errors)
else:
# Slight performance gain breaking dump into separate
# threads if > 1 replication target
for dm in self.doc_managers:
t = threading.Thread(target=do_dump, args=(dm, errors))
dumping_threads.append(t)
t.start()
# cleanup
for t in dumping_threads:
t.join()
# Print caught exceptions
try:
while True:
LOG.critical(
"Exception during collection dump", exc_info=errors.get_nowait()
)
dump_success = False
except queue.Empty:
pass
if not dump_success:
err_msg = "OplogThread: Failed during dump collection"
effect = "cannot recover!"
LOG.error("%s %s %s" % (err_msg, effect, self.oplog))
self.running = False
return None
if dump_cancelled[0]:
LOG.warning(
"Initial collection dump was interrupted. "
"Will re-run the collection dump on next startup."
)
return None
return timestamp
def dump_collection(self):
"""Dumps collection into the target system.
This method is called when we're initializing the cursor and have no
configs i.e. when we're starting for the first time.
"""
dump_set = self.namespace_set or []
logging.debug("OplogThread: Dumping set of collections %s " % dump_set)
#no namespaces specified
if not self.namespace_set:
db_list = retry_until_ok(self.main_connection.database_names)
for database in db_list:
if database == "config" or database == "local":
continue
coll_list = retry_until_ok(
self.main_connection[database].collection_names)
for coll in coll_list:
if coll.startswith("system"):
continue
namespace = "%s.%s" % (database, coll)
dump_set.append(namespace)
timestamp = util.retry_until_ok(self.get_last_oplog_timestamp)
if timestamp is None:
return None
long_ts = util.bson_ts_to_long(timestamp)
def docs_to_dump():
for namespace in dump_set:
logging.info("OplogThread: dumping collection %s"
% namespace)
database, coll = namespace.split('.', 1)
last_id = None
attempts = 0
# Loop to handle possible AutoReconnect
while attempts < 60:
target_coll = self.main_connection[database][coll]
if not last_id:
cursor = util.retry_until_ok(
target_coll.find,
fields=self._fields,
sort=[("_id", pymongo.ASCENDING)]
)
else:
cursor = util.retry_until_ok(
target_coll.find,
{"_id": {"$gt": last_id}},
fields=self._fields,
sort=[("_id", pymongo.ASCENDING)]
)
try:
for doc in cursor:
if not self.running:
raise StopIteration
doc["ns"] = self.dest_mapping.get(
namespace, namespace)
doc["_ts"] = long_ts
last_id = doc["_id"]
yield doc
break
except pymongo.errors.AutoReconnect:
attempts += 1
time.sleep(1)
# Extra threads (if any) that assist with collection dumps
dumping_threads = []
# Did the dump succeed for all target systems?
dump_success = True
# Holds any exceptions we can't recover from
errors = queue.Queue()
try:
for dm in self.doc_managers:
# Bulk upsert if possible
if hasattr(dm, "bulk_upsert"):
logging.debug("OplogThread: Using bulk upsert function for"
"collection dump")
# Slight performance gain breaking dump into separate
# threads, only if > 1 replication target
if len(self.doc_managers) == 1:
dm.bulk_upsert(docs_to_dump())
else:
def do_dump(error_queue):
all_docs = docs_to_dump()
try:
dm.bulk_upsert(all_docs)
except Exception:
# Likely exceptions:
# pymongo.errors.OperationFailure,
# mongo_connector.errors.ConnectionFailed
# mongo_connector.errors.OperationFailed
error_queue.put(sys.exc_info())
t = threading.Thread(target=do_dump, args=(errors,))
dumping_threads.append(t)
t.start()
else:
logging.debug("OplogThread: DocManager %s has not"
"bulk_upsert method. Upserting documents "
"serially for collection dump." % str(dm))
num = 0
for num, doc in enumerate(docs_to_dump()):
if num % 10000 == 0:
logging.debug("Upserted %d docs." % num)
dm.upsert(doc)
logging.debug("Upserted %d docs" % num)
# cleanup
for t in dumping_threads:
t.join()
except Exception:
# See "likely exceptions" comment above
errors.put(sys.exc_info())
# Print caught exceptions
try:
while True:
klass, value, trace = errors.get_nowait()
dump_success = False
traceback.print_exception(klass, value, trace)
except queue.Empty:
pass
if not dump_success:
err_msg = "OplogThread: Failed during dump collection"
effect = "cannot recover!"
logging.error('%s %s %s' % (err_msg, effect, self.oplog))
self.running = False
return None
return timestamp
def dump_collection(self):
"""Dumps collection into the target system.
This method is called when we're initializing the cursor and have no
configs i.e. when we're starting for the first time.
"""
timestamp = util.retry_until_ok(self.get_last_oplog_timestamp)
if timestamp is None:
return None
long_ts = util.bson_ts_to_long(timestamp)
dump_set = self.namespace_set or []
LOG.debug("OplogThread: Dumping set of collections %s " % dump_set)
# No namespaces specified
if not self.namespace_set:
db_list = retry_until_ok(self.primary_client.database_names)
for database in db_list:
if database == "config" or database == "local":
continue
coll_list = retry_until_ok(
self.primary_client[database].collection_names)
for coll in coll_list:
# ignore system collections
if coll.startswith("system."):
continue
# ignore gridfs collections
if coll.endswith(".files") or coll.endswith(".chunks"):
continue
namespace = "%s.%s" % (database, coll)
dump_set.append(namespace)
def docs_to_dump(namespace):
database, coll = namespace.split('.', 1)
last_id = None
attempts = 0
# Loop to handle possible AutoReconnect
while attempts < 60:
target_coll = self.primary_client[database][coll]
if not last_id:
cursor = util.retry_until_ok(
target_coll.find,
projection=self._projection,
sort=[("_id", pymongo.ASCENDING)]
)
else:
cursor = util.retry_until_ok(
target_coll.find,
{"_id": {"$gt": last_id}},
projection=self._projection,
sort=[("_id", pymongo.ASCENDING)]
)
try:
for doc in cursor:
if not self.running:
raise StopIteration
last_id = doc["_id"]
yield doc
break
except (pymongo.errors.AutoReconnect,
pymongo.errors.OperationFailure):
attempts += 1
time.sleep(1)
def upsert_each(dm):
num_inserted = 0
num_failed = 0
for namespace in dump_set:
for num, doc in enumerate(docs_to_dump(namespace)):
if num % 10000 == 0:
LOG.debug("Upserted %d docs." % num)
try:
mapped_ns = self.dest_mapping.get(namespace, namespace)
dm.upsert(doc, mapped_ns, long_ts)
num_inserted += 1
except Exception:
if self.continue_on_error:
LOG.exception(
"Could not upsert document: %r" % doc)
num_failed += 1
else:
raise
LOG.debug("Upserted %d docs" % num_inserted)
if num_failed > 0:
LOG.error("Failed to upsert %d docs" % num_failed)
def upsert_all(dm):
try:
for namespace in dump_set:
mapped_ns = self.dest_mapping.get(namespace, namespace)
dm.bulk_upsert(docs_to_dump(namespace), mapped_ns, long_ts)
except Exception:
if self.continue_on_error:
LOG.exception("OplogThread: caught exception"
" during bulk upsert, re-upserting"
" documents serially")
upsert_each(dm)
else:
raise
def do_dump(dm, error_queue):
try:
# Dump the documents, bulk upsert if possible
if hasattr(dm, "bulk_upsert"):
LOG.debug("OplogThread: Using bulk upsert function for "
"collection dump")
upsert_all(dm)
else:
LOG.debug(
"OplogThread: DocManager %s has no "
"bulk_upsert method. Upserting documents "
"serially for collection dump." % str(dm))
upsert_each(dm)
# Dump GridFS files
for gridfs_ns in self.gridfs_set:
db, coll = gridfs_ns.split('.', 1)
mongo_coll = self.primary_client[db][coll]
dest_ns = self.dest_mapping.get(gridfs_ns, gridfs_ns)
for doc in docs_to_dump(gridfs_ns + '.files'):
gridfile = GridFSFile(mongo_coll, doc)
dm.insert_file(gridfile, dest_ns, long_ts)
except:
# Likely exceptions:
# pymongo.errors.OperationFailure,
# mongo_connector.errors.ConnectionFailed
# mongo_connector.errors.OperationFailed
error_queue.put(sys.exc_info())
# Extra threads (if any) that assist with collection dumps
dumping_threads = []
# Did the dump succeed for all target systems?
dump_success = True
# Holds any exceptions we can't recover from
errors = queue.Queue()
if len(self.doc_managers) == 1:
do_dump(self.doc_managers[0], errors)
else:
# Slight performance gain breaking dump into separate
# threads if > 1 replication target
for dm in self.doc_managers:
t = threading.Thread(target=do_dump, args=(dm, errors))
dumping_threads.append(t)
t.start()
# cleanup
for t in dumping_threads:
t.join()
# Print caught exceptions
try:
while True:
LOG.critical('Exception during collection dump',
exc_info=errors.get_nowait())
dump_success = False
except queue.Empty:
pass
if not dump_success:
err_msg = "OplogThread: Failed during dump collection"
effect = "cannot recover!"
LOG.error('%s %s %s' % (err_msg, effect, self.oplog))
self.running = False
return None
return timestamp
def dump_collection(self):
"""Dumps collection into the target system.
This method is called when we're initializing the cursor and have no
configs i.e. when we're starting for the first time.
"""
timestamp = retry_until_ok(self.get_last_oplog_timestamp)
if timestamp is None:
return None
long_ts = util.bson_ts_to_long(timestamp)
# Flag if this oplog thread was cancelled during the collection dump.
# Use a list to workaround python scoping.
dump_cancelled = [False]
def get_all_ns():
ns_set = []
gridfs_ns_set = []
db_list = retry_until_ok(self.primary_client.database_names)
for database in db_list:
if database == "config" or database == "local":
continue
coll_list = retry_until_ok(
self.primary_client[database].collection_names)
for coll in coll_list:
# ignore system collections
if coll.startswith("system."):
continue
# ignore gridfs chunks collections
if coll.endswith(".chunks"):
continue
if coll.endswith(".files"):
namespace = "%s.%s" % (database, coll)
namespace = namespace[:-len(".files")]
if self.namespace_config.gridfs_namespace(namespace):
gridfs_ns_set.append(namespace)
else:
namespace = "%s.%s" % (database, coll)
if self.namespace_config.map_namespace(namespace):
ns_set.append(namespace)
return ns_set, gridfs_ns_set
dump_set, gridfs_dump_set = get_all_ns()
LOG.debug("OplogThread: Dumping set of collections %s " % dump_set)
def docs_to_dump(from_coll):
last_id = None
attempts = 0
projection = self.namespace_config.projection(from_coll.full_name)
# Loop to handle possible AutoReconnect
while attempts < 60:
if last_id is None:
cursor = retry_until_ok(
from_coll.find,
projection=projection,
sort=[("_id", pymongo.ASCENDING)]
)
else:
cursor = retry_until_ok(
from_coll.find,
{"_id": {"$gt": last_id}},
projection=projection,
sort=[("_id", pymongo.ASCENDING)]
)
try:
for doc in cursor:
if not self.running:
# Thread was joined while performing the
# collection dump.
dump_cancelled[0] = True
raise StopIteration
last_id = doc["_id"]
yield doc
break
except (pymongo.errors.AutoReconnect,
pymongo.errors.OperationFailure):
attempts += 1
time.sleep(1)
def upsert_each(dm):
num_failed = 0
for namespace in dump_set:
from_coll = self.get_collection(namespace)
mapped_ns = self.namespace_config.map_namespace(namespace)
total_docs = retry_until_ok(from_coll.count)
num = None
for num, doc in enumerate(docs_to_dump(from_coll)):
try:
dm.upsert(doc, mapped_ns, long_ts)
except Exception:
if self.continue_on_error:
LOG.exception(
"Could not upsert document: %r" % doc)
num_failed += 1
else:
raise
if num % 10000 == 0:
LOG.info("Upserted %d out of approximately %d docs "
"from collection '%s'",
num + 1, total_docs, namespace)
if num is not None:
LOG.info("Upserted %d out of approximately %d docs from "
"collection '%s'",
num + 1, total_docs, namespace)
if num_failed > 0:
LOG.error("Failed to upsert %d docs" % num_failed)
def upsert_all(dm):
try:
for namespace in dump_set:
from_coll = self.get_collection(namespace)
total_docs = retry_until_ok(from_coll.count)
mapped_ns = self.namespace_config.map_namespace(
namespace)
LOG.info("Bulk upserting approximately %d docs from "
"collection '%s'",
total_docs, namespace)
dm.bulk_upsert(docs_to_dump(from_coll),
mapped_ns, long_ts)
except Exception:
if self.continue_on_error:
LOG.exception("OplogThread: caught exception"
" during bulk upsert, re-upserting"
" documents serially")
upsert_each(dm)
else:
raise
def do_dump(dm, error_queue):
try:
LOG.debug("OplogThread: Using bulk upsert function for "
"collection dump")
upsert_all(dm)
if gridfs_dump_set:
LOG.info("OplogThread: dumping GridFS collections: %s",
gridfs_dump_set)
# Dump GridFS files
for gridfs_ns in gridfs_dump_set:
mongo_coll = self.get_collection(gridfs_ns)
from_coll = self.get_collection(gridfs_ns + '.files')
dest_ns = self.namespace_config.map_namespace(gridfs_ns)
for doc in docs_to_dump(from_coll):
gridfile = GridFSFile(mongo_coll, doc)
dm.insert_file(gridfile, dest_ns, long_ts)
except:
# Likely exceptions:
# pymongo.errors.OperationFailure,
# mongo_connector.errors.ConnectionFailed
# mongo_connector.errors.OperationFailed
error_queue.put(sys.exc_info())
# Extra threads (if any) that assist with collection dumps
dumping_threads = []
# Did the dump succeed for all target systems?
dump_success = True
# Holds any exceptions we can't recover from
errors = queue.Queue()
if len(self.doc_managers) == 1:
do_dump(self.doc_managers[0], errors)
else:
# Slight performance gain breaking dump into separate
# threads if > 1 replication target
for dm in self.doc_managers:
t = threading.Thread(target=do_dump, args=(dm, errors))
dumping_threads.append(t)
t.start()
# cleanup
for t in dumping_threads:
t.join()
# Print caught exceptions
try:
while True:
LOG.critical('Exception during collection dump',
exc_info=errors.get_nowait())
dump_success = False
except queue.Empty:
pass
if not dump_success:
err_msg = "OplogThread: Failed during dump collection"
effect = "cannot recover!"
LOG.error('%s %s %s' % (err_msg, effect, self.oplog))
self.running = False
return None
if dump_cancelled[0]:
LOG.warning('Initial collection dump was interrupted. '
'Will re-run the collection dump on next startup.')
return None
return timestamp
def dump_collection(self):
"""Dumps collection into the target system.
This method is called when we're initializing the cursor and have no
configs i.e. when we're starting for the first time.
"""
dump_set = self.get_dump_set()
shard_doc_count = self.shard_doc_count()
timestamp = None
LOG.info("OplogThread: Dumping set of collections %s from shard %s with %d docs" % (dump_set, self.shard_name, shard_doc_count))
if shard_doc_count > 0:
timestamp = util.retry_until_ok(self.get_last_oplog_timestamp)
long_ts = util.bson_ts_to_long(timestamp)
else:
long_ts = 0
def query_handle_id(query):
if '_id' in query and query['_id']:
for key in query['_id']:
query['_id'][key] = ObjectId(query['_id'][key])
def docs_to_dump(namespace):
database, coll = namespace.split('.', 1)
last_id = None
attempts = 0
# Loop to handle possible AutoReconnect
while attempts < 60:
target_coll = self.primary_client[database][coll]
fields_to_fetch = None
if 'include' in self._fields:
fields_to_fetch = self._fields['include']
query = {}
if self.initial_import['query']:
query_handle_id(self.initial_import['query'])
query = self.initial_import['query']
if not last_id:
cursor = util.retry_until_ok(
target_coll.find,
query,
fields=fields_to_fetch,
sort=[("_id", pymongo.ASCENDING)]
)
else:
cursor = util.retry_until_ok(
target_coll.find,
query.update({"_id": {"$gt": last_id}}),
fields=fields_to_fetch,
sort=[("_id", pymongo.ASCENDING)]
)
try:
for doc in cursor:
if not self.running:
LOG.error("Stopped iterating over Cursor while initial import")
raise StopIteration
self.pop_excluded_fields(doc)
last_id = doc["_id"]
yield doc
break
except (pymongo.errors.AutoReconnect,
pymongo.errors.OperationFailure):
attempts += 1
time.sleep(1)
# Extra threads (if any) that assist with collection dumps
dumping_threads = []
# Did the dump succeed for all target systems?
dump_success = True
# Holds any exceptions we can't recover from
errors = queue.Queue()
if len(self.doc_managers) == 1:
do_dump(self.doc_managers[0], errors)
else:
# Slight performance gain breaking dump into separate
# threads if > 1 replication target
for dm in self.doc_managers:
t = threading.Thread(target=do_dump, args=(dm, errors))
dumping_threads.append(t)
t.start()
# cleanup
for t in dumping_threads:
t.join()
# Print caught exceptions
try:
while True:
LOG.critical('Exception during collection dump',
exc_info=errors.get_nowait())
dump_success = False
except queue.Empty:
pass
if not dump_success:
err_msg = "OplogThread: Failed during dump collection"
effect = "cannot recover!"
LOG.error('%s %s %s' % (err_msg, effect, self.oplog))
self.running = False
return None
else:
LOG.info('OplogThread: Successfully dumped collection')
LOG.warning('Fields with errors: %r' % self.error_fields)
if not self.reimporting:
LOG.info("Waiting for other threads to finish collection dump")
self.barrier.wait()
return timestamp
def dump_collection(self):
"""Dumps collection into the target system.
This method is called when we're initializing the cursor and have no
configs i.e. when we're starting for the first time.
"""
dump_set = self.get_dump_set()
shard_doc_count = self.shard_doc_count()
timestamp = None
long_ts = 0
LOG.info("OplogThread: Dumping set of collections %s from shard %s with %d docs" % (dump_set, self.shard_name, shard_doc_count))
if shard_doc_count > 0:
timestamp = util.retry_until_ok(self.get_last_oplog_timestamp)
if timestamp:
long_ts = util.bson_ts_to_long(timestamp)
def query_handle_id(query):
if '_id' in query and query['_id']:
for key in query['_id']:
query['_id'][key] = ObjectId(query['_id'][key])
def docs_to_dump(namespace):
database, coll = namespace.split('.', 1)
last_id = None
attempts = 0
# Loop to handle possible AutoReconnect
while attempts < 60:
target_coll = self.primary_client[database][coll]
fields_to_fetch = None
if 'include' in self._fields:
fields_to_fetch = self._fields['include']
query = {}
if self.initial_import['query']:
query_handle_id(self.initial_import['query'])
query = self.initial_import['query']
if not last_id:
cursor = util.retry_until_ok(
target_coll.find,
query,
fields=fields_to_fetch,
sort=[("_id", pymongo.ASCENDING)]
)
else:
cursor = util.retry_until_ok(
target_coll.find,
query.update({"_id": {"$gt": last_id}}),
fields=fields_to_fetch,
sort=[("_id", pymongo.ASCENDING)]
)
try:
for doc in cursor:
if not self.running:
LOG.error("Stopped iterating over Cursor while initial import")
raise StopIteration
self.pop_excluded_fields(doc)
last_id = doc["_id"]
yield doc
break
except (pymongo.errors.AutoReconnect,
pymongo.errors.OperationFailure):
attempts += 1
time.sleep(1)
def upsert_all_failed_docs(dm, namespace, errors):
if errors:
for _id, field in errors:
self.upsert_doc(dm, namespace, long_ts, _id, field)
def upsert_each(dm):
num_inserted = 0
num_failed = 0
for namespace in dump_set:
mapped_ns = self.dest_mapping.get(namespace, namespace)
dm.disable_refresh(mapped_ns)
for num, doc in enumerate(docs_to_dump(namespace)):
if num % 10000 == 0:
LOG.info("Upserted %d docs." % num)
try:
self.upsert_doc(dm, namespace, long_ts, None, None, doc)
num_inserted += 1
except Exception:
if self.continue_on_error:
LOG.exception("Could not upsert document: %r" % doc)
num_failed += 1
else:
raise
dm.enable_refresh(mapped_ns)
LOG.info("Upserted %d docs" % num_inserted)
if num_failed > 0:
LOG.error("Failed to upsert %d docs" % num_failed)
def upsert_all(dm):
try:
for namespace in dump_set:
mapped_ns = self.dest_mapping.get(namespace, namespace)
dm.disable_refresh(mapped_ns)
errors = dm.bulk_upsert(docs_to_dump(namespace), mapped_ns, long_ts)
upsert_all_failed_docs(dm, namespace, errors)
dm.enable_refresh(mapped_ns)
except Exception:
if self.continue_on_error:
LOG.exception("OplogThread: caught exception"
" during bulk upsert, re-upserting"
" documents serially")
upsert_each(dm)
else:
raise
def do_dump(dm, error_queue):
try:
# Dump the documents, bulk upsert if possible
if hasattr(dm, "bulk_upsert"):
LOG.debug("OplogThread: Using bulk upsert function for "
"collection dump")
upsert_all(dm)
else:
LOG.debug(
"OplogThread: DocManager %s has no "
"bulk_upsert method. Upserting documents "
"serially for collection dump." % str(dm))
upsert_each(dm)
# Dump GridFS files
for gridfs_ns in self.gridfs_set:
db, coll = gridfs_ns.split('.', 1)
mongo_coll = self.primary_client[db][coll]
dest_ns = self.dest_mapping.get(gridfs_ns, gridfs_ns)
for doc in docs_to_dump(gridfs_ns + '.files'):
gridfile = GridFSFile(mongo_coll, doc)
dm.insert_file(gridfile, dest_ns, long_ts)
except:
# Likely exceptions:
# pymongo.errors.OperationFailure,
# mongo_connector.errors.ConnectionFailed
# mongo_connector.errors.OperationFailed
error_queue.put(sys.exc_info())
# Extra threads (if any) that assist with collection dumps
dumping_threads = []
# Did the dump succeed for all target systems?
dump_success = True
# Holds any exceptions we can't recover from
errors = queue.Queue()
if len(self.doc_managers) == 1:
do_dump(self.doc_managers[0], errors)
else:
# Slight performance gain breaking dump into separate
# threads if > 1 replication target
for dm in self.doc_managers:
t = threading.Thread(target=do_dump, args=(dm, errors))
dumping_threads.append(t)
t.start()
# cleanup
for t in dumping_threads:
t.join()
# Print caught exceptions
try:
while True:
LOG.critical('Exception during collection dump',
exc_info=errors.get_nowait())
dump_success = False
except queue.Empty:
pass
if not dump_success:
err_msg = "OplogThread: Failed during dump collection"
effect = "cannot recover!"
LOG.error('%s %s %s' % (err_msg, effect, self.oplog))
self.running = False
return None
else:
LOG.info('OplogThread: Successfully dumped collection')
LOG.warning('Fields with errors: %r' % self.error_fields)
if not self.reimporting:
LOG.info("Waiting for other threads to finish collection dump")
self.barrier.wait()
return timestamp
