Background
We have received feedback from a few people (I recall hguerrer_jira and others) having start-up problems with AMP 2.0, related to the different rate of start-up of different components, which combined with dependencies between them, can lead to unreliable and/or slow start-up of AMP overall.

Here is a recent example from rhn-support-nmasse who has spent considerable time looking at it, and has suggestions:

_Ordered startup of 3scale components. If I just use aml.yml to provision 3scale, it takes an undefined time to start ranging from 15min to 1 hour. Depending, on timeouts, retries, resource availability, etc. With the correct order, the startup time is around 1 minute.

Note: once the first startup is passed, subsequent restart are fine. But it seems that the database schema creation (that happens on first start) delays things a lot. _

Discussion
So, first I'd like to encourage some open, constructive discussion about what could be done to make it more reliable, and once reliable, quicker (i.e. speed is a secondary goal IMO).

So, being explicit about goals (based on previous discussions with Nicolas):
#1 reliable start-up (discarding some "below minimal resources" environments we're not targetting)

#2 faster start-up

#3 better / easier understanding of failures

#4 easier understanding of dependencies between components (mainly for goal #3... )