Uncovering Community and Project Insights Through Data Driven Methods
summary
Open source pipelines collect GitHub, Prow testgrid and Bugzilla data into Ceph S3 and a Trino-backed SQL store, crunch it in Jupyter notebooks and surface KPIs in Apache Superset while Kubeflow automates recurring jobs. Built ML services include a time-to-merge classifier (0–3 h, 3–6 h, 6–24 h, 1–7 days) trained on openshift/origin via srcopsmetrics and served with Seldon on OpenShift, and an optimal stopping point for long-running tests that flags likely-to-fail runs; models are exposed as APIs and packaged as a GitHub bot/action.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:45Agenda
- 1:32Why Metrics
- 13:39Metric Processing
- 15:19Examples of Machine Learning Models
- 16:42The Time To Merge Prediction Model
- 19:13Feature Engineering
- 21:01Next Steps
- 22:01Resources
related talks
31mTowards an open operations data community
34mUncovering Community and Project Insights Through Data Driven Methods
34mOptimize Long Test Runtimes Using Open AIOps
20mBuild an e2e analytics application using DataHub
36mData Science Meets DevOps: Gitops with OpenShift
33mCompliant Data Management and Machine Learning on Kubernetes
34mData pipelines powered by Open source and fun
35mBuilding and Scaling Data Pipelines for Data Engineering With Ceph Object Storage