feature: Cloudwatch alarms based on the Swarm state

[seriousben]

Is there a way to get cloudwatch alarms based on the Swarm state?

Some alarms that would be useful:

• Swarm worker terminated
• Swarm manager terminated
• Swarm managers not able to have quorum
• Swarm managers falling back behind in keeping up with the Gossip
• Swarm managers are not able to handle the load
• Some Swarm managers are forming their own Swarm on their side
• Swarm not able to schedule new tasks/services

It seems like the current way to do this is using prometheus + alert manager. But AFAIK this still doesn't cover most of the cases. It would be great, if as part of docker-for-aws, it integrated with cloudwatch alarms (bonus points with integrating with cloudwatch metrics)

[hairyhenderson]

Maybe this can be implemented by way of the new metrics plugin support added in 17.06? (moby/moby#32874)