Skip to content
tenement tenement

Fleet Mode (slum)

Experimental. slum exists as a library (~950 lines) but is not yet production-ready. The API described here works but will change.

What is slum?

Section titled “What is slum?”

tenement runs on a single server. slum coordinates multiple tenement instances across a fleet of servers, so you can distribute tenants geographically or run them with high availability.

+------------------------------------------+
|     slum (fleet database + router)       |
+-------------+-------------+-------------+
|  tenement   |  tenement   |  tenement   |
|  (east-1)   |  (west-1)   |  (south-1)  |
+-------------+-------------+-------------+

slum is a Rust library (not a standalone binary yet). You register servers, assign tenants to them, and slum tracks where everything lives in a SQLite database.

The idea

Section titled “The idea”

The core API is straightforward. You add servers, add tenants with a home server assignment, and spawn/stop instances across the fleet:

let db = SlumDb::init("slum.db").await?;


db.add_server(&Server {
    id: "east".into(),
    url: "http://east.example.com".into(),
    region: Some("us-east".into()),
    capacity_mb: 1024,
    ..Default::default()
}).await?;


db.add_tenant(&Tenant {
    domain: "customer-1.example.com".into(),
    server_id: "east".into(),
    process: "api".into(),
    instance_id: "prod".into(),
    ..Default::default()
}).await?;


db.spawn_instance("customer-1", "api", "prod", "east").await?;

What’s next

Section titled “What’s next”

The main planned integration is with haqlite for high-availability tenants. The idea is that two tenement servers + S3 gives you HA without Kubernetes: haqlite handles SQLite WAL replication to S3, and slum handles failover (detecting a dead server and spawning tenants on the surviving one).

This would make the single-server limitation soft rather than hard. You’d still write single-tenant code and deploy with tenement, but your tenants would survive a server failure.

Current status

Section titled “Current status”

slum handles server registration, tenant assignment, and instance orchestration across servers. It enforces referential integrity (can’t delete a server with active tenants, tenant’s server must exist). What it doesn’t do yet is automatic failover, health monitoring of servers, or geographic routing. Those are planned.

Next steps

Section titled “Next steps”