Two repositories, two containers, one pipeline that deploys them together and a rollback that has to move both.
the tests, and the answers:
can one be deployed without the other? no
can one be rolled back independently? no
do they share a database schema? yes
can they be released on different days? no
does either have a separate team? no
five noes. this is one service with a network hop in the
middle, and the hop costs 4ms and an entire class of
failure.The distribution is real and the independence is not, which is the worst combination — all of the operational cost of two services and none of the benefit. Merging them back was a fortnight and removed a retry policy, a circuit breaker and a shared client library. The lesson worth keeping is that the boundary test is deployment independence, not repository count.