Currently, scap3 uses a list of targets defined in the scap directory of the repo being deployed. This list of targets may become out-of-date with the currently-pooled list of machines for a service.
Ops shouldn't have to maintain this list in every repo. Deployers should still have the ability to determine how best to group servers for a particular services deployment.
In the deployment tooling meeting today we discussed having a global list of pooled servers per service (already exists) against which the list of all targets of the current deploy could be checked.
If there is a difference master list vs. repo, the deploy in progress could then alert the deployer of the situation and/or do something sane.
The current plan of attack here is to attempt to get a list of servers directly from conftool and skip the intermediate step of generating dsh lists for MediaWiki.
Unknowns here (@Joe may be able to answer these):