New Docker container runtime health check is not resilient

[captainsafia]

The new Docker container runtime health check (docker buildx version) is not sufficient since it will return with a zero exit-code even if Docker is not running. We need to figure out a way to validate if Docker is running and supports buildx.

[danegsta]

DCP runs docker container ls -n 1 to list the latest container if one exists; it's fast unless Docker is in resource saver mode in which case you run into the usual problem of the command potentially going unresponsive.

[captainsafia]

it's fast unless Docker is in resource saver mode in which case you run into the usual problem of the command potentially going unresponsive.

Does this happen with docker info as well?

DCP runs docker container ls -n 1 to list the latest container if one exists

I like the idea of unifying the approach we use between DCP and the AppHost. For this particular case, there is an additional requirement to check that BuildKit is installed as well. I suppose it is fine to have the health check spin up two separate processes for each.

[danegsta]

We've moved away from running docker info. Even when not in resource saver mode we've seen many examples where the command could take the better part of a minute to run.