Interested in what processes you all have encounte...
# general
j
Interested in what processes you all have encountered to stay on top of dependencies in a large organization.
a
which types, for systems loosely couple, use forcing functions to validate (DiRTs, failovers etc)
have strong invariants that you can test on a regular basis
n
From an SOA perspective (this service calls this service) a developer portal like Backstage can be part of the solution, but depends on active maintenance of the dependencies. I haven't seen this done well in any size of organization 😂. Following for interest!
a
So the Bazel dependency graph is another way. It will be correct and if you enforce a few standards around rpc system it will be 100% correct. For example if you use grpc + proto as your interface you have your graph. We are tackling this on the prod side at prodvana.io in deploy as well in a similar way. I think the IDP solution is kinda BS. It doesn’t create real back pressure to have the correct graph.
n
Yeah, anything that uses code or standards to deterministically create the graph is 💯 better than the "depend on humans to annotate stuff"
k
For a platform with not-well-implemented tracing we were asked to migrate to AWS --> we simply used config-maps + plantUML 🙂 but it was a one-time job, only for a migration planning. But it's a very specific scenario. In my opinion, Platform Team should closely work with Enterprise Architecture