Since you're using EE I recommend contacting docker support for that, definitely sounds like a misconfiguration. success.docker.com/support

What instance sizes are you using? Don't use T2 or anything that cap's CPU. Also be sure you're sizing resources based on UCP/DTR requirements and best practice arch. docs.docker.com/datacenter/ucp/2.2/guides/admin/install/… success.docker.com/article/docker-ee-best-practices

Other useful tools could be docker events, docker service inspect, and docker container inspect (on that ify7vhl0l2m8). I'd also do a watch docker service ps serviceNameX to see if tasks were re-spawing but indeed it's weird to see a single task "running" but a 0/1 in service ls.

OK I updated my answer to ensure you're testing the networks in the way they are intended, and also corrected my statement about 17.06 vs. 17.09.

Hmm, is IPv6 enabled on Windows (it needs to be). Can you curl a Linux nginx container from Windows? The availability of ping working in VIP's has been different across versions so I wouldn't rely on it as a connectivity test.

While Peter is correct, the idea you have about desktops in a dynamic swarm has lots of concerns and potential issues in its design. Swarm managers need to be an odd number and generally only 3-7 of them in a Swarm. Nodes must have well-connected networks to each other with a list of ports open and no NAT between them. Your idea sounds fun, but is not what Swarm was designed for unless your desktops are all Linux hosts that mostly stay up and managers are on fixed IP's.