Swarm watch to local non-TLS etcd proxy mode with TLS etcd cluster fail

[qkboy]

For security reason,I built three nodes TLS-enabled etcd cluster as swarm/docker kv backend.
But the libnetwork plugin just can connect to non-TLS etcd service.

So I follow this reference :https://coreos.com/etcd/docs/latest/etcd-live-http-to-https-migration.html.

If you typically connect to a remote etcd cluster, this is a good time to configure an etcd proxy that handles the remote connection and TLS termination, and to reconfigure your apps to communicate through the proxy on localhost.

I startup a local non-TLS etcd proxy as those service to use.
When I run swarm join or manage service , the first connection is sucess, the second and others are all error.

swarm manage:

[root@docker-node01 ~]# ./swarm manage -H :2476 --advertise docker-node01:2476 --replication --discovery-opt kv.disable_autosync=true --tlsverify --tlscacert=/etc/docker/certs.d/ca.pem --tlscert=/etc/docker/certs.d/docker-node01-member.pem --tlskey=/etc/docker/certs.d/docker-node01-member-key.pem etcd://localhost:2379              INFO[0000] Initializing discovery without TLS           
INFO[0000] Listening for HTTP                            addr=:2476 proto=tcp
INFO[0000] Leader Election: Cluster leadership lost     
INFO[0000] New leader elected: docker-contiv02.idc.vipshop.com:2476 
INFO[0000] Registered Engine docker-node02 at docker-node02:2376 
INFO[0000] Registered Engine docker-node01 at docker-node01:2376 
ERRO[0003] Discovery error: Unexpected watch error      
ERRO[0005] client: etcd cluster is unavailable or misconfigured 
ERRO[0005] Leader Election: watch leader channel closed, the store may be unavailable...

swarm join:

# ./swarm join --advertise=docker-node02:2376 etcd://localhost:2379
INFO[0000] Initializing discovery without TLS           
INFO[0000] Registering on the discovery service every 1m0s...  addr=docker-node02:2376 discovery=etcd://localhost:2379
INFO[0060] Registering on the discovery service every 1m0s...  addr=docker-node02:2376 discovery=etcd://localhost:2379
ERRO[0060] client: etcd cluster is unavailable or misconfigured 
INFO[0120] Registering on the discovery service every 1m0s...  addr=docker-node02:2376 discovery=etcd://localhost:2379
ERRO[0120] client: etcd cluster is unavailable or misconfigured 

I think it may something like this issue:
etcd-io/etcd#3894

So how can I use like etcdctl --no-sync options in swarm ?
THX.