How to Monitor Application Performance in a Container-Based World
summary
Cisco demonstrates a concrete monitoring stack for Mantl/Kubernetes that collects cAdvisor, collectd and load‑balancer (traffic) metrics, converts them to Monasca-style metrics with four labels (projectID, serviceID, environment, host), and streams everything over a Kafka measurement bus into InfluxDB. The pipeline enables inline alarming (alarms evaluated on write, before DB persistence), correlates response‑time spikes with container/host metrics, and drives automated scale up/down actions, affinity changes or suspended execution for compromised hosts.
AI-written summary · about
chapters (12)
- 0:00Start
- 0:42Problem statement (requirement)
- 1:26Motivations
- 9:11Solution Pillars • Unified Measurements Format
- 11:54Measurement Bus and Services
- 12:23Visualization using Grafana
- 14:14Compound Measurements
- 15:00Adding Load Balancer (Traefik)
- 16:01Load Balancer Metrics
- 16:48Adding Security Scans (Cisco Norad)
- 18:27Complete Solution
- 19:21Takeaways
related talks
50mMonitoring and metrology for containers
46mMonitoring and metrology for containers
42mDistributed Monitoring: How to understand the Chaos
29mKubernetes-Defined Monitoring [I]
28mLightweight Observability with Performance Co-Pilot and Grafana
42mUnified Monitoring of Containers and Microservices [I]
42mMonitoring Mesos with InfluxDB & the TICK Stack
49mComprehensive Container-Based Service Monitoring with Kubernetes and Istio