fscl-process-svc

Bounded-context service crate for the process view.

Kubernetes naming uses process-api for the workload to avoid colliding with the Kubernetes Service suffix, but the git repo and crate name stay fscl-process-svc.

The crate currently compiles in an earlier design state. It is not yet migrated to use ComponentLifecycleUow end to end, and its future NATS-consumer role is scaffolded at the deployment layer rather than fully implemented in-process.

Dependencies

fscl-process-svc
    |
    +--> fscl_core
    |
    +--> migration

dev / k8s runtime
    |
    +--> sidecar fscl-outbox-publisher

And the deployment-level relationship is:

process-api container  --> PostgreSQL
process-api container  --> NATS consumer role (scaffolded)
outbox-publisher sidecar --> PostgreSQL outbox
outbox-publisher sidecar --> NATS publisher role

Split

fscl-process-svc        -> HTTP API, process view persistence, current app logic
fscl_core               -> shared domain/application contracts
fscl-messaging          -> shared outbox/wire contract, used indirectly today
fscl-outbox-publisher   -> sidecar publisher runtime

Config

Shared/local dev loading order:

1. ../.env.shared
2. local .env
3. container or shell overrides

Current local config points:

• DB_TYPE: database scheme, currently postgres
• DB_HOST: database host
• DB_PORT: database port
• DB_USER: database user
• DB_PASSWORD: database password
• DB_NAME: process bounded-context database name
• APP_HOST: bind address for the HTTP server
• APP_PORT: bind port for the HTTP server
• NATS_URL: reserved for the process-api consumer role and shared local setup
• NATS_JETSTREAM_STREAM: target JetStream stream name for the bounded context
• NATS_JETSTREAM_DURABLE_CONSUMER: durable consumer name for process-api
• NATS_JETSTREAM_ACK_POLICY: intended acknowledgement policy
• NATS_JETSTREAM_ACK_WAIT: intended acknowledgement timeout

Today the DB settings are used by the running crate. The NATS consumer settings are scaffolding for the target architecture.

Dev Setup

Create the env files:

cp ../.env.shared.example ../.env.shared
cp .env.example .env

Run tests/build locally:

cd fscl-process-svc
cargo test
cargo run

Run the local dev stack:

docker compose -f ../compose/infra.yaml -f ../compose/process-stack.yaml up

The service currently applies its own SeaORM migration on startup.

K8s Setup

Scaffold only for now.

Relevant manifests live under fscl/doc/rust/fscl-k8s:

• 15-process-messaging.yaml: shared messaging runtime values for the bounded context
• 20-process-api.yaml: process-api deployment and sidecar wiring
• 22-outbox-publisher.yaml: publisher-local settings
• 30-postgres.yaml: bounded-context database

Target workload split:

process-api pod
  |- process-api container       -> serves HTTP, will consume NATS events
  |- outbox-publisher sidecar    -> relays DB outbox rows to NATS

The database-owning service remains responsible for applying both its own schema and the shared outbox schema.