import sys
import time
import datetime
import logging,logging.handlers
import exceptions
import pymongo
import bson
from pymongo import ReplaceOne
import myconfig as config
class MongoSynchronizer(object):
""" MongoDB multi-source synchronizer.
"""
def __init__(self, src_hostportstr, dst_hostportstr, **kwargs):
""" Constructor.
"""
if not src_hostportstr:
raise Exception('src hostportstr is empty')
if not dst_hostportstr:
raise Exception('dst hostportstr is empty')
self._src_hostportstr = src_hostportstr
self._dst_hostportstr = dst_hostportstr
self._src_mc = None
self._dst_mc = None
self._start_optime = None # if true, only sync oplog
self._last_optime = None # optime of the last oplog has been replayed
self._logger = logging.getLogger()
self._start_optime = kwargs.get('start_optime')
self._src_mc = pymongo.MongoClient(src_hostportstr)
self._dst_mc = pymongo.MongoClient(dst_hostportstr)
def __del__(self):
""" Destructor.
"""
if self._src_mc:
self._src_mc.close()
if self._dst_mc:
self._dst_mc.close()
def _sync(self):
""" Sync databases and oplog.
"""
if self._start_optime:
self._logger.info("locating oplog, it will take a while")
oplog_start = bson.timestamp.Timestamp(int(self._start_optime), 0)
doc = self._src_mc['local']['oplog.rs'].find_one({'ts': {'$gte': oplog_start}})
if not doc:
self._logger.error('specified oplog not found')
return
oplog_start = doc['ts']
self._logger.info('start timestamp is %s actually' % oplog_start)
self._last_optime = oplog_start
self._sync_oplog(oplog_start)
def _sync_oplog(self, oplog_start):
""" Replay oplog.
"""
try:
host, port = self._src_mc.address
self._logger.info('try to sync oplog from %s on %s:%d' % (oplog_start, host, port))
cursor = self._src_mc['local']['oplog.rs'].find({'ts': {'$gte': oplog_start}}, cursor_type=pymongo.cursor.CursorType.TAILABLE, no_cursor_timeout=True)
#Tailable cursors are only for use with capped collections.
if cursor[0]['ts'] != oplog_start:
self._logger.error('%s is stale, terminate' % oplog_start)
return
except IndexError as e:
self._logger.error(e)
self._logger.error('%s not found, terminate' % oplog_start)
return
except Exception as e:
self._logger.error(e)
raise e
self._logger.info('replaying oplog')
n_replayed = 0
while True:
try:
if not cursor.alive:
self._logger.error('cursor is dead')
raise pymongo.errors.AutoReconnect
# get an oplog
oplog = cursor.next()
# guarantee that replay oplog successfully
recovered = False
while True:
try:
if recovered:
self._logger.info('recovered at %s' % oplog['ts'])
recovered = False
self._replay_oplog(oplog)
n_replayed += 1
if n_replayed % 1000 == 0:
self._print_progress(oplog)
break
except pymongo.errors.DuplicateKeyError as e:
# TODO
# through unique index, delete old, insert new
#self._logger.error(oplog)
#self._logger.error(e)
break
except pymongo.errors.OperationFailure as e:
# TODO
# through unique index, delete old, insert new
#self._logger.error(oplog)
#self._logger.error(e)
break
except pymongo.errors.AutoReconnect as e:
self._logger.error(e)
self._logger.error('interrupted at %s' % oplog['ts'])
self._dst_mc = pymongo.MongoClient(self._dst_hostportstr)
if self._dst_mc:
recovered = True
except pymongo.errors.WriteError as e:
self._logger.error(e)
self._logger.error('interrupted at %s' % oplog['ts'])
self._dst_mc = pymongo.MongoClient(self._dst_hostportstr)
if self._dst_mc:
recovered = True
except StopIteration as e:
# there is no oplog to replay now, wait a moment
time.sleep(0.1)
except pymongo.errors.AutoReconnect:
self._src_mc.close()
self._src_mc = pymongo.MongoClient(self._src_hostportstr)
cursor = self._src_mc['local']['oplog.rs'].find({'ts': {'$gte': self._last_optime}}, cursor_type=pymongo.cursor.CursorType.TAILABLE, no_cursor_timeout=True)
if cursor[0]['ts'] != self._last_optime:
self._logger.error('%s is stale, terminate' % self._last_optime)
return
def _replay_oplog(self, oplog):
self._replay_oplog_mongodb(oplog)
def _print_progress(self, oplog):
ts = oplog['ts']
self._logger.info('sync to %s, %s' % (datetime.datetime.fromtimestamp(ts.time), ts))
def _replay_oplog_mongodb(self, oplog):
# Replay oplog on destination if source is MongoDB.
# parse
ts = oplog['ts']
op = oplog['op'] # 'n' or 'i' or 'u' or 'c' or 'd'
ns = oplog['ns']
dbname = ns.split('.', 1)[0]
if op == 'i': # insert
collname = ns.split('.', 1)[1]
self._dst_mc[dbname][collname].insert_one(oplog['o'])
#self._dst_mc[dbname][collname].replace_one({'_id': oplog['o']['_id']}, oplog['o'], upsert=True)
elif op == 'u': # update
collname = ns.split('.', 1)[1]
self._dst_mc[dbname][collname].update(oplog['o2'], oplog['o'])
elif op == 'd': # delete
collname = ns.split('.', 1)[1]
self._dst_mc[dbname][collname].delete_one(oplog['o'])
elif op == 'c': # command
self._dst_mc[dbname].command(oplog['o'])
elif op == 'n': # no-op
pass
else:
self._logger.error('unknown operation: %s' % oplog)
self._last_optime = ts
def run(self):
""" Start data synchronization.
"""
# never drop database automatically
# you should clear the databases manually if necessary
try:
self._sync()
except exceptions.KeyboardInterrupt:
self._logger.info('terminating')
def logger_init(filepath):
""" Init logger for global.
"""
logger = logging.getLogger()
logger.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s [%(levelname)s] %(message)s')
if filepath:
handler_log = logging.handlers.RotatingFileHandler(filepath, mode='a', maxBytes=1024*1024*100, backupCount=3)
handler_log.setFormatter(formatter)
handler_log.setLevel(logging.INFO)
logger.addHandler(handler_log)
else:
handler_stdout = logging.StreamHandler(sys.stdout)
handler_stdout.setFormatter(formatter)
handler_stdout.setLevel(logging.INFO)
logger.addHandler(handler_stdout)
def main():
logger_init(config.cf.get('conf','log'))
syncer = MongoSynchronizer(
config.cf.get('db','mongo_f_uri'),
config.cf.get('db','mongo_t_uri'),
start_optime=config.cf.get('conf','start_time'))
syncer.run()
if __name__ == '__main__':
main()
7547
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
