swarm doesn't recover if its been turned off for 2 months.

[SvenDowideit]

I just visited someone who's raspberry pi based swarm thing failed to come back after being off for 2 months.

eventually i noticed in docker system info that swarm was complaining that the cert had expired

running docker node or docker stack ls results in an error that was similar to "swarm was never initialised"

but docker ps -a clearly showed containers with service task names

i deleted the cert files, and then docker was very convinced that swarm had never been initialised

running docker swarm init got us to an uninitialised swarm with no stacks or services

and then docker deploy... brings things back, and seems to use the abandoned swarm created and managed volumes

so I'm hoping that the only fix needed, is that swarm sees that the cert is out of date, and replaces it (this is a one node swarm, and swarmkit can probably detect that this is the case?)

extra weirdness is that the existance of these certs seems to be the only thing swarm's using to determine if there are stacks and services it at some point had - that seems like a shortcut that could probably be improved on.

that said, it works,