Unable to Configure MongoDB Replication Without Initial Sync

database-designmongodbmongodumpreplication

I have a big MongoDB database on my production server, which I want to set up as a replica set. I understand that if I enable the replica set option on both servers, it will automatically start to sync the data from scratch. But I don't want to sync from the beginning; instead I want to configure it like MySQL replication.

I would like to follow the following process:

  • Take dump from primary
  • Restore it on secondary
  • Enable replica set in config file, on both servers
  • Restart MongoDB.
  • Go to primary server and add the secondary node.

Is this possible; or is there any other alternate solution for this?

Best Answer

Why do two sets of heavy lifting ??? Let MongoDB do it.

For this example, suppose IP of SECONDARY is 10.30.50.70 with hostname myslave

STEP 01 : Enable Replica Set on the PRIMARY only

Make the following changes in /etc/mongodb.conf on the PRIMARY

  • enable authorization in /etc/mongodb.conf
    • authorization: enabled under the security YAML tag (MongoDB 2.6+)
    • auth=true (if you are using MongoDB 2.4)
  • Enable replication (suppose your name the replica set myreplica)
    • Put replSetName: myreplica under replication YAML tag (MongoDB 2.6+)
    • Put replSet = myreplica (MongoDB 2.4)

STEP 02 : Restart MongoDB on the PRIMARY

service mongod restart

STEP 03 : Initiate Replica Set on the PRIMARY

rs.initiate();

NOTE: At this point your have a one-node Replica Set

STEP 04 : Perform Steps 01 and 02 on your SECONDARY

NOTE: At this point, the SECONDARY is replica set enabled but not replicating

STEP 05 : NOW, add the SECONDARY to the PRIMARY

Goto the PRIMARY and run

rs.add("10.30.50.70:27017")

or

rs.add("myslave:27017")

STEP 06 : Check status of Replica Set

Run this repeatedly

rs.status()

If this command ever freezes, don't worry. It rebuilding all the indexes.

You can open another ssh session on the SECONDARY, run tail -f on the mongodb.log file and see the index rebuild progression.

WHY DO IT THIS WAY ???

MongoDB will do essentially start replication, copy the data, and rebuild indexes in a single operation. As shown, this process is in no way automatic when building an initial replica set. Doing it this way, is less error-prone.

Related Question