This repository includes scripts and playbooks to test and benchmark Ansible at scale.
This is targeted at profiling for Kayobe and Kolla Ansible, but the insights will be generally useful for large scale Ansible use. No real hosts are used - only local tasks are benchmarked. This allows us to test how inventories of various sizes affect Ansible controller performance.
Documentation and results of benchmarks are provided here.
profile parser parses the output of the profile_tasks callback plugin and generates CSV data. It's not pretty, but it is effective if this is the only data you have access to.
profile massager massages the CSV data generated by the profile parser, into various formats. It aggregates based on task duration, and allows filtering by play, role or task state.
Apache 2.0
- Mark Goddard (mark@stackhpc.com)