Multiple managers in swarm cluster does not respond

I am creating my first swarm cluster. By now, I have 3 nodes with rancherOS, virtualized on Vmware vsphere.
With one single manager node, the cluster is working right. But I want to add a manager node for HA.
Whan I convert a worker node to a manager node, the cluster makes it unworkable. I have delay on nodes console, and I receive erors of this type:

kernel:[ 6869.402953] watchdog: BUG: soft lockup - CPU#0 stuck for 24s! [ntpd:1017]

If I convert the second manager to worker, the cluster works fine again.

Some ideas.