Sentin8:Captain — Process Supervisor
The starter that runs the fleet: health-gated tier spawning (the DAG
of boot order), rolling zero-downtime upgrades (upgrade-instance:
start new → wait Active → drain old → stop), remote spawn across
machines, placement enforcement (singletons, edge pins), and machine
provisioning backends (docker/Proxmox/AWS).
Use cases
- Fleet bring-up —
make startboots captain, which spawns the enrolled platform + vertical services in dependency tiers. - Rolling upgrades — a new binary ships to one instance at a time
while the proxy round-robins traffic; drain happens only after the
successor registers
Active. - Fleet placement — singleton services (acme, proxy) enforce one-instance; edge services pin to the edge machine; the autoscaler spawns remotely when saturation demands.
- Provisioning — tofu kits (docker/Proxmox/cloud) grow the fleet on demand, with graceful-leave deprovisioning.
Operator surface
| Command | Role |
|---|---|
make status | fleet status, KEY=<ip:port> per instance |
make upgrade-instance KEY=… | rolling zero-downtime swap |
make restart-instance KEY=… | hard restart of one instance |
make stop/start-instance KEY=… | stop/start one instance |
Captain's own REST surface (POST /v1/captain/instances,
POST /v1/captain/shutdown, health) is certificate-guarded.
Full documentation
The complete detail lives in this product's sub-pages:
- Architecture (DDD) — the full document, harvested from the unit's repository at build time.
- Business rules (BRD) — the full document, harvested from the unit's repository at build time.
- Security — the full document, harvested from the unit's repository at build time.
- Overview (repo README) — the maintainer's README in the consolidated reference section.