Docker-machine is not cleaning up after run

[mathuin]

So I’ve made great strides with my setup which means I’ve found a new problem.

I have WebODM using ClusterODM on DigitalOcean for processing with autoscaling, and it successfully spun up a new droplet via docker-machine. The new droplet finished the task within an hour or so, but it’s still sticking around. The ClusterODM README mentions spin up/down so I figure it should be doing the thing. I poked about a bit in the source code and the only time I see it destroying machines is when tasks are aborted. Am I missing something?

[pierotofy]

You probably need to set --public-address when starting ClusterODM: ClusterODM/config.js at master · OpenDroneMap/ClusterODM · GitHub

ClusterODM tries to infer it, but it’s often wrong. At the end of processing, nodes make a callback to that address to tell ClusterODM they are done (and tear down the machine).

[smathermather]

Setting public address has fixed most idiosyncrasies I’ve seen with clusterodm.

[mathuin]

Thank you for the tip! The primary droplet and the spun-up droplets are in the same VPC so I can use the internal IP for this purpose, and I’m 95% sure that’s exposed by the time I spin up the clusterodm container.