steps:

• do_pre_check,
  • disable balancer
  • src/dest count
  • print src/dest chunks
• do_getlastop,
  • check if oplog is large enough for this migration
  • save lastop timestamp
• do_export_parallel,
  • export, chunk by chunk
• do_shard,
  • enable sharding for dest db/collection.
  • pre-split and moveChunk (for balance in dest cluster)
• do_copyindex,
  • copy index
• do_import_parallel,
  • import, chunk by chunk.
  • we can use a safe version of mongoimport, and retry
• do_oplogreplay,
  • replay oplog, from lastop.
• do_post_check,
  • enable balancer again
  • src/dest count
  • print src/dest chunks

• pymongo: https://github.com/mongodb/mongo-python-driver
• pcl: https://github.com/idning/pcl

conf/ms_conf.py::

#mongo binary dir
MONGO_PATH = '/home/ning/bin'

#set a name for sharding clusters 
shard_online_0 = ('10.23.250.156:7333', 'user', 'passwd')
shard_offline_0 = ( '10.65.19.52:37333', 'user', 'passwd')

::

./bin/ms.py --src shard_online_0 --dest shard_offline_0 --src_ns db.collection --dest_ns db.collection -v --stop_balancer

if something bad intercept the oplogreplay, we can just replay oplog from some timestamp::

./bin/ms.py --src shard_online_0 --dest shard_offline_0 --src_ns db.collection --dest_ns db.collection -v --replay_from 1379127546

::

--worker 5