We have selections of docker containers that run as a client on a host remotely.
This then communicates with a single server that has a front-end and is in charge of multiple of these clusters.
We'd like to implement the ability to remotely tell a cluster to self destruct, from the server.
However, since the clients run in docker containers, it's tough to ensure all traces of their data and volumes are removed.
My thinking so far is to use one top-level docker-in-docker container with an API the server can call that kills all the other containers.
Is there an alternative to this I could look into? Another requirement is that this is very reliable, which I fear may not be the case if some containers cannot gracefully stop.
Many thanks!
