I’ve installed latest docker release candidate from the testing repo
(currently = Docker version 1.12.0-rc3, build 91e29e8)
Created a swarm of (eg) 4 nodes, promote three to manager (for a consensus)
Created a service (standard nginx) with scale 6 replicas.
I find that not all instances work. On perhaps one or two nodes, the containers will be created attempting to assign an IP address that’s aready in use (within the internal docker swarm IP range), so the containers won’t start. The new container is left with status “created”. Docker realises the container is not started so creates a new instance, again attempting to use the same IP address. This process continues ad-infinitum. When the created containers reach ten on a node, docker removes the oldest one (and continues creating new ones).
Has anyone else seen this problem?