Using etcd-io/raft library

06-two-node-snapshot

In this section, we let one node of the cluster receive the full FSM state from another node via a snapshot + remaining entries.

This requires implementing the remaining steps of the "state machine handling loop":

• Applying snapshots to FSM, and notifying raft library via ReportSnapshot() (link).
• Remembering ConfChange returned from ApplyConfChange() (link).

The test 06-two-node-snapshot/the_test.go is similar to the previous 05-two-node-cluster, except before creating the second node, it performs a log compaction on node 1. So that when node 2 joins the cluster, it has to recover the FSM state from the snapshot.

Note that the Snapshot is not coming from a "committed" channel. When Ready struct contains a Snapshot, we apply it to both MemoryStorage and to the FSM.

Next: 07-detour-change-fsm.

This site is open source. Improve this page.