Cindy noted at our architecture meeting today that the configuration file at https://github.com/GPII/universal/blob/master/gpii/configs/gpii.config.cloudBased.production.json which claims to hold our production cloudbased flowManager configuration has some problematic contents.
For a start, it derives from a development configuration, which is reported at https://issues.gpii.net/browse/GPII-1056 to be a faulty practice. In particular, it inherits a definition of a built-in preferences server, which isn't appropriate for a production configuration. It seems that when deployed in "real production", part of this configuration is overridden to point to the production preferences server - which is problematic, since the purpose of these configurations is to be complete and usable in the JSON form in which they sit.
Some discussion is written up at the end of today's architecture notes at http://lists.gpii.net/pipermail/architecture/2016-August/004268.html
It looks like the deployment structure is here:
These yaml files need to be replaced by a scheme like that deployed privately in the "first discovery server": https://github.com/GPII/first-discovery-server/blob/master/src/js/configUtils.js
The test cases for this config are somewhat inadequate, but they do at least verify that it starts up and functions as a flow manager. There is some long standing CI work by Alfredo to fire up two VMs holding a flowManager and a preferences server that I'm not sure of the status of.
As a wider issue, it is hard to answer questions of this kind, for those who are not members of the ops team. I asked today, "How can someone discover what public-facing servers are deployed, and what branches and what configuration they are built from", to which the answer currently pointed to a private IDRC github repository - that is, one which can't be read by non-members of the IDRC ops team. There is a JIRA open to move this to the GPII, but I think the wider issue will remain.
It should be possible to answer this question from the QI infrastructure via some kind of dashboard.