Configuring Replication Sets

A replication set config doc...

• is JSON formatted
• shared across all nodes in a set
• can manually set changes to doc with...
  • rs.add
  • rs.initiate
  • rs.remove
  • rs.reconfig
  • rs.config

Config Highlights

{
  _id: <string>,
  version: int,
  members: [...]
}

• _id: matches the server-defined replica set. the --replSet identifier in the mongod clu. AND/OR the replication: replSetName: value in the mongod.conf file. These values in these locations cannot conflict.
• version: increments when the replicaSet changes
  • like when a new node gets added, or removed, when the topology of the replicaSet
• members: holds "topology" of the replica set
  • array of objs, each obj representing a member of the replica set
    • _id: node id
    • host: hostname:port
    • arbiterOnly: bool //false by default, the role of the node
    • hidden: bool // not visible to app. rs.isMaster WONT find this node. helpful for handling reporting
    • priority: int // 0 - 1000. higher priorities get elected more often. setting priority to 0 pretty much makes sure a member will never become primary. Change in priority triggers new election. When a member is an arbiterOnly, the priority needs to be 0. When a node is hidden, the priority must be 0.
    • slaveDelay: int //delay in seconds, default to 0. When set to, say 3600 (1 hour), this member replicates data from other nodes 1-hour-ago. This implies hidden/0-priority values. Why?! Hmm.

There are a lot more details to configure: settings, version, configsvr... these are beyond basic.

Final Thoughts

• replication config doc defines the replica set: properties are defined and shared across the set
• members field notes deets about each replSet member