Skip to main content

One Server, Multiple Clients

CourseMaker is not multi-tenant — every client gets their own fully isolated instance. That's a deliberate product decision: each client owns their Stripe account, their Bunny library, their data.

If you operate CourseMaker as a hosting provider for several clients on the same physical machine, run one stack per client:

/opt/coursemaker/client-acme/
  docker-compose.yml
  .env                   # NEXT_PUBLIC_APP_URL=https://learn.acme.com
                         # APP_HOST_PORT=4210

/opt/coursemaker/client-beta/
  docker-compose.yml
  .env                   # NEXT_PUBLIC_APP_URL=https://learn.beta.com
                         # APP_HOST_PORT=4211

Each compose project gets a unique name: (top of the compose file) so volumes and networks don't collide:

name: coursemaker-acme

Reverse proxy each public hostname to its own APP_HOST_PORT.

Resource sizing​

Per stack, expect:

  • 200–400 MB RAM idle for the app container.
  • 100–200 MB RAM idle for Postgres.
  • < 1 % CPU at idle, spikes during catalogue browsing or migration runs.

A 4 GB / 2 vCPU VPS comfortably hosts 4–6 small instances. Above that, isolate per-VPS so a runaway instance can't starve its neighbours.

Backups across multiple stacks​

Loop over each /opt/coursemaker/client-* directory in the cron job and dump each Postgres separately. Each stack has its own ENCRYPTION_KEY — back up keys per-client.