Meet with the Vault team, key stakeholders, and one or two application teams
(current users or onboarding candidates).
Collect data on a set of onboarding candidates: application ID, platform, runtime,
language, and where it runs (pipelines, Kubernetes, cloud, on-prem).
Identify the prevailing client personas and the viable auth method for each.
Scope which secret types and usage to support first.
Define the namespace strategy: root reserved for platform administration,
namespaces only for business-unit or operating-team boundaries, ACL policies for
team-level isolation.
Select two or three consumption patterns, prioritized by highest impact and lowest
effort.
Agree on the onboarding request format (YAML or HCL) and the approval flow.
- Build the onboarding automation
Build a Terraform module per consumption pattern that turns a handful of high-level
inputs into the full Vault configuration (roles, policies, secret paths).
Stand up a central onboarding repo in Azure DevOps with documentation and a
request template. Any team can open a pull request, so teams on other CI platforms
don't need to run the modules themselves.
Wire approval to a Terraform pipeline that provisions access automatically.
Use Terraform templating to generate a welcome message: login URL, required Okta
group, secret path, a seeded example secret, a pipeline snippet, and links to docs
and support.
Onboard a small cohort of willing teams that covers each pattern, and gather
feedback on what works and what isn't clear.
Expand to more teams with documentation only, treating them as beta testers.
Graduate to fully self-service pull requests, opening access gradually.
- Migrate and drive adoption
Route all new onboardings to the shared-namespace model while existing teams
keep running where they are.
Communicate the new model internally and set a re-onboarding window (around 12
months) for existing teams, with self-service tooling such as KV secret migration.
Page 3 of 4