MongoDB supports different high availability configurations; master-slave replication, Replica Set and Sharded Cluster. The most popular deployment we see is Replica Set.
Replica Set has a number of advantages:
- Multiple copies of data (if a server breaks down, the data is available on another server)
- Failover (if the primary fails, a secondary can take over)
- Potentially extra read performance if you explicitly allow secondaries to serve reads
- Possible deployment on two nodes (with arbiter running on a separate host)
In this post, we are going to show on how to install ClusterControl on top of your existing MongoDB/TokuMX Replica Set using the bootstrap script. Note that you are able to colocate ClusterControl with any of the mongod instaces.
Requirement
Prior to this installation, whether it is Replica Set or Sharded Cluster, make sure your MongoDB or TokuMX instances meet the following requirements:
- ClusterControl requires every mongod instance to have an associate PID file. This includes shardsvr, configsvr, mongos, and arbiter. To define a PID file, use --pidfilepath variable or pidfilepath option in the configuration file. Starting from version 1.2.5, ClusterControl is able to detect PID even without specifying pidfilepath.
- All MongoDB binaries must be installed on each node identically under one of the following paths:
- /usr/bin
- /usr/local/bin
- /usr/local/mongod/bin
- /opt/local/mongodb
- MongoDB Replica Set/Sharded Cluster has been configured as a cluster. Verify this with sh.status() or rs.status() command.
- Ensure that the designated ClusterControl node meets the hosts requirement.
Architecture
We have a three-node MongoDB Replica Set (one primary, one secondary and one arbiter) running on Ubuntu 12.04. All mongod instances have been installed through 10gen apt repository. ClusterControl controller will be installed on mongo3 so this is where we will perform the installation. mongo3 is configured as an arbiter to the replica set.
