check, based on counting bytes of RPCs, that each command is sent to each peer just once.

TestReElection测试在上一个测试的基础上，检查发生leader离线情况的行为。 先等待集群产生一个leader，然后让这个leader离线，检查集群能否再产生一个leader。 若能，令原来的leader上线，检查leader不应该发生变化，因为后来的leader的term大于原来的le

regression: an RPC that's delayed during Enabled=false should not delay subsequent RPCs (e.g. after Enabled=true).

do the servers persist the snapshots, and restart using snapshot along with the tail of the log?

if one server falls behind, then rejoins, does it recover by using the InstallSnapshot RPC? also checks that majority discards committed log entries

are the snapshots not too huge? 500 bytes is a generous bound for the operations we're doing here.

test static 2-way sharding, without shard movement.

test unreliable

applier reads message from apply ch and checks that they match the log contents

periodically snapshot raft state

必须注册才能进行解码和编码

as each Raft peer becomes aware that successive log entries are committed, the peer should send an ApplyMsg to the service (or tester) on the same se

主要是applyCh的处理中，ApplyMsg的Command是一个interface，因此要向labgob注册具体实现才能进行编解码

Partition servers into 2 groups and put current leader in minority

Sets up 2 partitions with connectivity between servers in each partition.

Randomize server handles