Beyond K8s Monitoring: Data to Knowledge to Action
summary
Kubernetes health metrics get pivoted into one-hot symptom rows, reduced with UMAP/PCA and clustered with DBSCAN, then each cluster gets Apriori/FP‑growth pattern mining. The pipeline surfaces symptom combinations prioritized by difference in percent occurrence versus the rest of the fleet, and matches many SME rules; it produces six new OTA rules from customer OpenShift data. Code and runnable notebooks are published, customer data withheld.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:21Preamble
- 7:15Pattern Mining
- 10:11Clustering
- 10:52Dimensionality Reduction
- 17:08How Do You Contribute and Get Involved
- 17:36Open Shift Anomaly Detection
related talks
22mHidden Signals in K8s Clusters
18mK8s Troubleshooting Demystified: Five Best Practices
37mProactive Kubernetes Security: Anomaly Detection and Runtime Alertin...
24mUncovering Community and Project Insights Through Data Driven Methods
11mKubernetes Observability
25mKube Intelligence - A Metric Based Insightful Remediation Recommender
29mKubernetes-Defined Monitoring [I]
30mBetter data quality with ML & visualization