WritingTests.md

Writing Tests With Maestro: Using

This guide extends the Using guide and focuses on the Maestro API used for testing.

Introduction

Maestro tests are fundamentally written around the Maestro class. The Maestro class wrap the maestro orchestration API and provides an interface that can be used to control the test cluster.

High Level Testing Logic

The tests have a pre-defined workflow that, no matter how complicated the test, can be summarized as having the following steps:

Therefore, the secret to write more complicated tests with Maestro is understanding what part of the API to call for each of those steps.

• Connect
  • Constructor
• Find Peers and Assign Roles
  • pingRequest
  • roleAssign
  • groupAssign (experimental)
• Set Test Parameters
  • setDuration
  • setParallelCount
  • setMessageSize
  • setMessageSize
  • setRate
  • setFCL
  • setManagementInterface
• Start the Test
  • startInspector
  • startAgent
  • startWorker
• Collect the Results
  • waitForNotifications
  • waitForDrain
• Cleanup
  • roleUnassign
  • groupLeave (experimental)

All of the core API methods return CompletableFutures that can be used to read the responses from the test cluster. The CompletableFutures return a list of responses, one for each node on the test cluster. The responses are guaranteed to be for the specific request since the API, as of 1.4 and newer, uses a correlation identifier to match the responses to the request that originated them.

Note: the response processing may occurs in two ways depending on the expected amount of responses. If a request is known to provide a fixed number of replies, then the collection of those replies will succeed if the given amount of replies matches to the expected value. If the amount is unknown, then it will wait for matching replies until it considers that the growth of matching replies has staled.