Autoscaling: Give feedback in the logs in case something wrong happens with a starting EC2

[sanderegg]

Currently the Autoscaling starts instances in AWS EC2 using:

• run_instances from the AWS API, passing UserData script to automatically join the swarm
• once the node joins the swarm it is part of the cluster and labelled as such

Since #3902 it is possible to modify the bash script (UserData) using an ENV variable.
Therefore errors (syntax, etc) might arise more often. It would be really good that the Autoscaling service shows some error in the logs to facilitate debugging such mistakes.

To this end it should monitor the state of the started machines and report in its logs when issues are detected.

[sanderegg]

maybe doing something such as

ssh autoscaled-node "cat /var/log/cloud-init-output.log"

would be possible?

[sanderegg]

Or as @mrnicegyu11 mentioned a maybe nicer way would be to shellcheck the bash script BEFORE sending it to the EC2 interface. This way we could validate the bash script.