Optimize Long Test Runtimes Using Open AIOps
summary
Open-source toolkit AI4CI collects GitHub Actions, Prow and TestGrid runs via Jupyter notebooks and Kubeflow pipelines, trains models on OpenShift, and surfaces KPIs. It builds an Optimal Stopping Point classifier that buckets test runtimes, computes failure rate per bucket and flags times where failure rate exceeds a 75% threshold so long-running outliers (most tests complete under a minute) can be terminated. A GitHub Action bot posts the prediction on PRs; accuracy depends on available training data.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:27AI for CI motivation
- 3:56Optimal stopping classifier
- 8:29Optimal stopping model
- 15:58Demo data collection
- 23:29Open source resources
- 26:20Q&A accuracy & data
- 34:15Closing
related talks
24mUncovering Community and Project Insights Through Data Driven Methods
44mWhat’s Breaking My Build? AIOps Tooling to Aid Your CI/CD Workflows
26mAIOps for CI with Kubeflow Pipelines
21mAI-driven optimization of CI/CD pipelines
29mAIOps Is How We Will Survive DevOps
30mGoing Plaid: Striving for Speed of Light in CI pipelines
01h 26mOptimising Continuous Delivery
30mPragmatic, scalable, and ergonomic CI/CD using Github Actions and K8s.