The following is what is used to benchmark "write to all members".
# write to all members
benchmark --endpoints=${HOST_1},${HOST_2},${HOST_3} --conns=100 --clients=1000
put --key-size=8 --sequential-keys --total=100000 --val-size=256
My understanding is that only the leader listens to write requests. So how is it possible to write to all members of an etcd cluster?
The secondary nodes just forward the write requests they receive to the leader.