I am attempting to use a TX1 as worker within a docker swarm.
The swarm also contains two x86_64 nodes - the master and another worker - both running stock latest revisions of Ubuntu 18.04 and docker-ce 18.09.3.
The TX1 is in a standard dev board, is running a somewhat stripped down install of L4T from Jetpack 3.3, and with docker-ce 18.09.3.
Using docker on the TX1 standalone works fine for simple things at least. I can start and stop containers and connect to them etc.
When I then try to add the TX1 to the cluster after the usual set of info level messages wrt the gossip cluster getting wired up I then see two error messages in ‘journalctl -u docker.service’:
Mar 19 10:06:47 tegra-ubuntu dockerd: time=“2019-03-19T10:06:47Z” level=error msg=“enabling default vlan on bridge br0 failed open /sys/class/net/br0/bridge/default_pvid: permission denied”
Mar 19 10:06:47 tegra-ubuntu dockerd: time=“2019-03-19T10:06:47.662790794Z” level=error msg=“reexec to set bridge default vlan failed exit status 1”
… which I don’t see on my other worker.
The TX1 appears to have joined the cluster successfully from the manager:
$ docker node ls
ID HOSTNAME STATUS AVAILABILITY MANAGER STATUS ENGINE VERSION
s14o76ap2sdgf2g7jfyka8b5h geoff-OldMacBook Ready Active 18.09.3
1k74jna5lhfba50ks6g7k0r7e tegra-ubuntu Ready Active 18.09.3
94zgws7ym9dwbo2x6be67hp9u * toc17-office Ready Active Leader 18.09.3
When I try to deploy a stack which puts a simple container onto the TX1 based on the following compose file fragment:
command: stdbuf -o L rostopic pub /turtle1/cmd_vel geometry_msgs/Twist -r 1 – ‘[2.0, 0.0, 0.0]’ ‘[0.0, 0.0, -1.8]’
constraints: [node.hostname == tegra-ubuntu]
It seems to get stuck in a continual fail-restart loop. If I deploy it to the other worker instead it works fine. I can run that container image directly on the TX1 outside the swarm via docker run
I am guessing that something in the L4T setup is causing the docker swarm overlay network not to be created correctly? Has anybody come across this before?