Learn how to identify hidden patterns in cluster behavior that impact reliability. Discover methods to detect issues like pod delays and service dependencies, using clustering algorithms and proactive strategies for improved resilience and resource optimization.
Join us for a deep dive session that extracts data-driven observations and insights from heavily utilized clusters to find unique patterns that impact reliability. We'll share how to implement the right signals to detect common challenges such as delays in pod spin-ups, pod termination data to identify workloads most affected by these terminations, or closely occurring failures that indicate hidden service dependencies. We'll then cover methods to preemptively address these to improve resilience using clustering algorithms like DBSCAN, for dependency mapping, and other good practices. We'll walk through a typical journey from experimental setups using scripts and command-line outputs to production-ready practices––from quick prototyping, to refining data pipelines and optimizing performance for operational environments. By uncovering these hidden patterns & signals you can achieve proactive reliability improvements, for improved cluster health & optimization of resource allocation.
Gary Edgar is a seasoned technical support and customer engineer at Komodor with 15+ years of experience across cloud infrastructure, databases, and monitoring systems. Currently at Komodor, he helps teams streamline Kubernetes troubleshooting.
Previously, Gary was a key early hire at PlanetScale, supporting major clients like Stripe and Twitter with performance tuning, architecture, and hands-on support. He also delivered impactful solutions during his time at SevOne (IBM), working onsite with Microsoft and eBay to improve system reliability and monitoring at scale.
