[functionbeat] Expose cloudformation error when waiting for status

[ph]

Currently, we only wait for Completed or Failed events for cloudformation; it's not the best thing for debugging. Instead, we should also monitor Stack Events and report them to the console as they appear.

This way it makes it a lot easier to debug why a stack failed to apply or not.

[ph]

To simplify writing other feature I am currently working on this, The DescribeStacksEvents is the API that you want to hit to retrieve any events related to a specific stack.

The problem is this API actually returns all the events for a stack, are you cannot say retrieve events after a specific date, so you actually have to do something like this.

• Fetch Events
• Move cursor to event for a resource type of stack
• Event must have one of the following status: CREATE_, UPDATE_ or DELETE_*
• We must assert the time to make sure its older than the beginning of our stack. (some jitter might be necessary)

When the following match this mean we have our first relevant event 🎉
Retrieve calls is a matter of doing multiple call over the API using the NextToken option to scope the pagination results.

BUT, When a stack is currently in progress nextToken can be nil because we have finished reading all the current events. So we also need to keep track of events that we have already received.

[ph]

I think the easiest way is to implement this:

• Have a goroutine that check only the status of a stack.
• Have a goroutine that only do the events reporting.

When a stack operation is detected to be complete by the first go routine we say to the second one to stop on next event.

[ph]

This seems to work pretty well, I have the full output of the cloudformation event, this mean when a create stack job fails I can do a nice cleanup job.