Service vs. Library
Title
Service vs. Library
Patlet
Teams in a DevOps environment may be reluctant to work across team boundaries on common code bases due to ambiguity over who will be responsible for responding to service downtime. The solution is to realize that often it's possible to either deploy the same service in independent environments with separate escalation chains in the event of service downtime or factor a lot of shared code out into one library and collaborate on that.
Problem
When teams are working in a DevOps environment developers are responsible for a feature end-to-end: From the customer down to deployment, maintenance and support. This poses a challenge when working across team boundaries: Escalation chains may not be the same for errors happening in either team. Coupling source code and deployment leaves the teams with the question of who is responsible for on-call support in the event of errors. As a result teams are reluctant to join forces even if there is significant overlap in requirements.
Context
Teams are working in a micro-services environment.
They are organized in fully functional DevOps teams: Each team is responsible for their contributions end-to-end, including maintenance, on-call and customer support.
A team is tasked with providing a service to their downstream customers that is fairly similar to an existing service built by another team.
Forces
Organizational escalation paths may be different for each of the teams.
Members of each team may be unwilling to answer on-call support for errors that do not affect their own downstream customers.
Severity levels for the same types of errors may be different across team boundaries due to different SLA definitions per team/customer relationship.
Teams may have different security or regulatory constraints governing their deployments.
Solutions
Decouple responsibility for source code from deployment: Both teams work to identify exactly where there is overlap and synergies.
Only shared source code is kept as part of the InnerSource project with shared responsibility. The shared source should be coherent in that it includes all testing code (including integration tests if available) and as much of the CI pipeline as is possible to make contribution validation easier.
Decouple configuration and deployment pipelines from actual business logic. Establish a second deployment of the service for the second team.
Treat the common base as a library that is used by both teams with shared code ownership.
Deployment configurations can be included as separate projects in your InnerSource portfolio to allow teams to discuss/collaborate or a new team to copy them.
Resulting Context
Teams are willing to collaborate, benefiting from sharing the work of implementing the business logic.
A service that originally was built specifically to work in one environment is converted into a more general solution based on a specific business requirement.
Both teams get to know their respective escalation policy and deployment setup, potentially identifying improvements for their own setup.
The likelihood that changes are needed and made in the shared source code increases, leading to more frequent opportunities to refine, improve and optimize the implementation.
Encourages incremental operational standardization in release packaging, telemetry, health/readiness endpoints and so on as the teams realize they can more efficiently maintain this in the shared code if they agree on standard conventions.
See also
Related to this pattern is the 30 Day Warranty pattern that takes a different approach to solving the forces described above.
Known Instances
Europace AG
Flutter Entertainment: A Flutter InnerSource application has a shared code "service" repository with cross-team contribution and CI pipeline to build and publish a shared release artifact. Each adopting team has a "deployment config" repository defining their own deployment. This is driven by varying regulatory requirements, service and incident management practices and infrastructure skill sets in different areas of the business.
WellSky (see Continuous InnerSource in Production - 5 Times)
Status
Structured
Author(s)
Isabel Drost-Fromm
Rob Tuley
Acknowledgments
Thank you Tobias Gesellchen for review internal to Europace AG.
Alias
Service vs. library: It's inner source, not inner deployment
Last updated
