Docker swarm on NVidia Jetson TX1 - bridge device issues

Docker Engine Swarm
2

Having dug a little further into it it does seem to be a network setup issue in the vicinity of the TX1.

If I create a quiescent container on each worker using the following compose fragment:

pub:
image: ros:melodic-ros-core
environment:
- “ROS_MASTER_URI=http://ros-master:11311”
- “ROS_HOSTNAME=pub”
command: sleep 999999
deploy:
placement:
constraints: [node.hostname == tegra-ubuntu]

pub2:
image: ros:melodic-ros-core
environment:
- “ROS_MASTER_URI=http://ros-master:11311”
- “ROS_HOSTNAME=pub2”
command: sleep 999999
deploy:
placement:
constraints: [node.hostname == geoff-OldMacBook]

… and then shell in to each by docker exec’ing /bin/bash I find that:

  • ros-master resolves to 10.0.22.8 on both workers
  • “ping ros-master” works correctly on both workers
  • “curl http://ros-master:11311” succeeds on the other worker but gets connection refused on the TX1

Intriguing!,

Geoff

PS. ros-master is another container in the swarm running on the manager node