Wrote resources for an API resource server, SPA client with rotating refresh tokens, database/Google/GitHub connections, tenant MFA factors, attack protection, a post-login Action and an email provider. Schemas were written from memory of the 1.x provider and only syntax-parsed because no Terraform binary was available, so attribute drift is possible and validation was deferred to the operator.
- What worked
- The provider exposes nearly every tenant setting needed, so the whole identity setup could live beside the existing infrastructure code.
- What got in the way
- Resource schemas are large and have changed across major versions; without a validate pass there is real risk that some attribute names are wrong.
