Using etcd-io/raft library

08-running-cluster

This section changes the cluster to have 3 nodes, and starts all nodes at once (as opposed to proposing ConfChanges to join the cluster). Again, we just follow the instructions from the etcd-io/raft README.

The test at 08-running-cluster/the_test.go:

• Creates a 3-node cluster (still running in the same process).
• Waits for one of them to be elected leader.
• Sends random proposals to random nodes.
• Has each node creating snapshots and compacting its log as it wishes.
• Lets it bake and verifies that all 3 nodes have the same FSM state at the end.

It also simulates an unreliable network by sometimes dropping messages addressed to a random target node.

Next: 09-async-storage-write.

This site is open source. Improve this page.