When a worker is killed, there is no way to recover it. We should be able to recover a worker if it is restarted with the exact same configuration.
State could likely just be pushed to etcd under some prefix when a worker is killed, and similarly a worker should check for existing state when it's started.
When a worker is killed, there is no way to recover it. We should be able to recover a worker if it is restarted with the exact same configuration.
State could likely just be pushed to etcd under some prefix when a worker is killed, and similarly a worker should check for existing state when it's started.