Skip to content

Latest commit

 

History

History
49 lines (29 loc) · 2.67 KB

README.md

File metadata and controls

49 lines (29 loc) · 2.67 KB
Raw

Provisioning

Ansible is used as a provisioning tool for the host running as controller. The playbook for that is located in playbooks/controller.yaml. The playbook targets the hosts under controllers in the ìnventory/inventory.yaml. To access the host it uses the ssh key referenced in ìnventory/group_vars/controller named ansible_ssh_private_key_file. The playbook also requires you to provide a password for the grafana (action) user.

Prerequisites

Install the required packages for your OS:

A typical way to run the playbook would be:

ansible-playbook -i infrastructure/inventory/inventory.yaml infrastructure/playbooks/controller.yaml --extra-vars "action_user_pw=test"

Architecture

The diagram below describes the production setup.

Production diagram

Production is comprised of two hosts:

  • The controller is a Virtual Machine with the runner, datastore and dashboard
  • The minion is a Bare metal machine to run the benchmarks.

Controller

The controller runs the runner as a daemon in a docker container and exposes an endpoint to be able to trigger a benchmark run with parameters.

Also the datastore, InfluxDB, and the dashboard, Grafana, are run inside containers, all using docker-compose.

The runner can be triggered from a remote resource but will, at least, be triggered once a day on a schedule to ensure there is a data point for each day.

The runner is responsible for running the latest benchmark tests on the minion. Ansible is used to manage the provisioning state of the minion before each run. The benchmarks write the test results in a file determined by a parameter. After the run, this file is retrieved and parsed by the runner. The results are then written to the datastore as well as stored on the IPFS network.

Minion

This bare metal machine will be kept as clean as possible to enable reliable results from the benchmarks. Tests are run over ssh from the controller.

After the benchmark test is run to get an accurate result the same test will be run using clinic tools to provide in depth analysis.

Tests kept up to date before each run and are always run in sequence. After each test, any generated files and resources will be retrieved by the runner and cleaned up after successful reception.