metrics

README

Metrics End to End Tests

This folder contains scripts used for triggering a load test on AWS. This is intended to be used before cutting a release to run the same set of workloads against two different versions of Nomad. It currently uses the end to end test terraform setup that relies on OSS build artifacts stored in AWS S3.

The current steps are:

1. Trigger a build on CircleCI:
   • Ensure that you have CIRCLE_API_TOKEN env var set. It's available as release-bot-token in https://circleci.hashicorp.engineering/gh/hashicorp/nomad-enterprise/edit#api . Use envchain or other tools if convenient
   • Run ./scripts/metric_runner.sh <nomad_oss_sha>
2. Examine the output from the final step of the CircleCI job. It should show URLs to monitor where you can check various metrics via Prometheus

Future Work

Standardize the set of metrics to use as a dashboard. This will require bringing Graphana or another graphing system into the test infrastructure managed by terraform.

Documentation

Index

• type MetricsTest
  • func (tc *MetricsTest) AfterAll(f *framework.F)
  • func (tc *MetricsTest) AfterEach(f *framework.F)
  • func (tc *MetricsTest) BeforeAll(f *framework.F)
  • func (tc *MetricsTest) TestMetricsLinux(f *framework.F)
  • func (tc *MetricsTest) TestMetricsWindows(f *framework.F)

Types

type MetricsTest

type MetricsTest struct {
	framework.TC
	// contains filtered or unexported fields
}

func (*MetricsTest) AfterAll

func (tc *MetricsTest) AfterAll(f *framework.F)

Clean up fabio/prometheus

func (*MetricsTest) AfterEach

func (tc *MetricsTest) AfterEach(f *framework.F)

Clean up the target jobs after each test case, but keep fabio/prometheus for reuse between the two test cases (Windows vs Linux)

func (*MetricsTest) BeforeAll

func (tc *MetricsTest) BeforeAll(f *framework.F)

Stand up prometheus to collect metrics from all clients and allocs, with fabio as a system job in front of it so that we don't need to have prometheus use host networking

func (*MetricsTest) TestMetricsLinux

func (tc *MetricsTest) TestMetricsLinux(f *framework.F)

TestMetricsLinux runs a collection of jobs that exercise alloc metrics. Then we query prometheus to verify we're collecting client and alloc metrics and correctly presenting them to the prometheus scraper.

func (*MetricsTest) TestMetricsWindows

func (tc *MetricsTest) TestMetricsWindows(f *framework.F)

TestMetricsWindows runs a collection of jobs that exercise alloc metrics. Then we query prometheus to verify we're collecting client and alloc metrics and correctly presenting them to the prometheus scraper.