中文wiki前往

This simpe program is for transfering data between two mongo nodes, including:

• data cloning;
• oplog syncing(applying one mongo databbase's oplog to another mongo database) ;
• oplog cloning(not applying, just storing the oplogs from source mongo database to another one)

• Get the source code

  $ git clone https://github.com/Qihoo360/mongosync.git

• Compile

  Enter the source code and typewrite:

  $ make

  after that, a "mongosync" named execution file appears in the $(MONGOSYNC_ROOT)/output directory, and also a "mongosync.conf" named config file

  SP:

  a. maybe some depenncies not installed in your machine, install them according to the tips（mostly sons, boost and ssl is needed）

  b. don't use c++11 to compile this program, otherwise will occur error

• Exectution

  The program can launches with command line or config file，and the specified usage can be obtained by:

  $ cd output
  $ ./mongosync --help
  Follow is the mongosync-surpported options:
  --help                   to get the help message
  -c conf.file             use config file to start mongosync
  --src_srv arg            the source mongodb server's ip port
  --src_user arg           the source mongodb server's logging user
  --src_passwd arg         the source mongodb server's logging password
  --src_auth_db arg        the source mongodb server's auth db
  --src_use_mcr            force source connection to use MONGODB-CR password machenism
  --dst_srv arg            the destination mongodb server's ip port
  --dst_user arg           the destination mongodb server's logging user
  --dst_passwd arg         the destination mongodb server's logging password
  --dst_auth_db arg        the destination mongodb server's auth db
  --dst_use_mcr            force destination connection to use MONGODB-CR password machenism
  --db arg                 the source database to be cloned
  --dst_db arg             the destination database
  --coll arg               the source collection to be cloned
  --dst_coll arg           the destination collection
  --oplog                  whether to sync oplog
  --raw_oplog              whether to only clone oplog
  --op_start arg           the start timestamp to sync oplog
  --op_end arg             the start timestamp to sync oplog
  --dst_op_ns arg          the destination namespace for raw oplog mode
  --no_index               whether to clone the db or collection corresponding index
  --filter arg             the bson format string used to filter the records to be transfered

Rough testing under the situation:

• Ping from this program's exectution's machine to the mongo db server's machine(two machines) is about 1.5ms；
• total about 36,400,000 documents；
• database size about 1.09GB

transfering time is about 379 seconds(even qps: 96.04k).

SP: The bottleneck is writing, So using multiple threads(multi-conn) to accelerate the speed of that(but also limited by database server's writing speed).