Provision and Orchestrate

OpenLMIS today uses docker compose to define a set of services that all work together in the ref-distro, as well as to define developer / testing environments for specific services or services to test together.

What we don't have:

• A way to deploy Ref Distro across multiple Docker Hosts. Being tied to one host limits how the host can scale (vertically and not horizontally to many hosts), and it results in a single point of failure for implementers. This also might be useful to developers if the entire thing can't fit on a laptop.

• A way to provision new cloud resources needed in launching / stopping instances.

This epic will evaluate some of the new tooling ecosystems that are emerging that co-provide orchestration and provisioning capabilities:

• Provision Docker hosts

• Combine hosts into a cluster

• Scale the cluster up and down as needed

  • possibility, though with less priority, to do this automatically in response to demands on the cluster

• Provision other cloud resources

  • Load balancers

  • Domain names

  • SSL Certs

  • Storage

This will mostly focus on tools such as Kubernetes and Kops, though others such as Ansible or Cloud Formation, AWS ECS or Lightsail could help fill in gaps / be considered.

Value: ability to scale development, QA, demo resources on demand, reduce overhead of creating and maintaining demo instances. Reference for implementers to help them speed configuring resources and software (LMIS configuration is out of scope though)