Arista ACE-S Automation (Foundations): Automation and Programmability Tooling

What the exam expects you to distinguish

Arista's programmability stack has several layers, and questions test whether you know which tool solves which problem.

• eAPI - JSON-RPC over HTTP(S) to EOS. You send runCmds (CLI commands) or get/enable/config methods. It is synchronous request/response and returns JSON. Good for ad-hoc automation and for tools that speak HTTP.
• pyeapi - a Python wrapper around eAPI. It gives you JSON output and connection management, but it is still CLI-driven underneath.
• RESTCONF / gNMI / OpenConfig - model-driven interfaces. gNMI is the streaming telemetry and config protocol; OpenConfig provides vendor-neutral YANG models.
• Ansible / Terraform / Python - orchestration and IaC layers that call the above.

Traps

• eAPI is not model-driven. It executes CLI commands; the output is structured JSON but the interface is command-based, not YANG-based. A question that says "eAPI uses OpenConfig models" is wrong.
• pyeapi is not a separate protocol. It wraps eAPI.
• gNMI is for streaming telemetry and model-based config, not for arbitrary CLI execution.
• Idempotency. Ansible modules and Terraform resources are expected to be idempotent; raw eAPI runCmds is not.

Worked example

To read interface counters from a switch in Python, you would use pyeapi: connect with pyeapi.connect_to("switch1"), then node.enable("show interfaces"). The same thing over raw HTTP is a POST to /command-api with {"jsonrpc":"2.0","method":"runCmds","params":{"cmds":["show interfaces"],"format":"json"}}. Knowing both shapes is the point.