Skip to main content

Going to production

Manual installation​

Orbits requires two persistent services:

  • MongoDB service
    Refer to the MongoDB documentation for setup details.
    Orbits stores a small amount of long-lived data in MongoDB. You may consider using MongoDB Atlas with a small cluster for hosting your database.

  • Node.js service running your action code
    See the Node.js documentation for production best practices.

  • 2 GB RAM and 1 CPU is generally sufficient; adjust based on your workload.
  • Deployment options for redundancy:
    • Kubernetes deployment with at least two pods.
    • ECS service with at least two tasks.

If your actions are not exposed via a web server, you do not need to expose your deployment through a service or load balancer.

Independent Worker Deployment​

The main goal in production is to ensure at least one Node.js process runs an ActionJob.

  • Control the number of ActionJob instances per Node.js process with the ORBITS_WORKERS__QUANTITY environment variable.
  • Example design:
    • Tasks with ORBITS_WORKERS__QUANTITY=0 (no workers), dedicated to hosting a web server or other purposes
    • Tasks with ORBITS_WORKERS__QUANTITY=2 dedicated solely to running ActionJob workers

Orbits cloud (β)​

With Orbits Cloud, you don’t need to manage the MongoDB cluster neither the Node.js service.

You also benefit from:

  • Automatic scaling of your Orbits services;
  • Automatic deployment from your Git repositories;
  • A frontend to visualize your agents and workflows;
  • Persistent logging;
  • Insights into your actions, workflows, and agents;
  • Permissions and secrets management over actions, workflows, and agents; If you’re interested in accessing the beta, click here.