Monitoring in the Microservices world: Kubernetes vs Serverless based applications
summary
Argues that Kubernetes requires monitoring at five layers (infrastructure, cluster, nodes, pods, application) using Prometheus/Alertmanager and Grafana, while serverless shifts responsibility to the cloud and forces tracing across managed services with correlation ID + OpenTracing/Jaeger or AWS CloudWatch and X‑Ray. Demonstrates a tool that auto-builds distributed traces across Lambda, S3, DynamoDB and shows per-request cost because serverless bills by milliseconds, plus open‑source Lambda cost analyzers on a public GitHub; demo UI has visibility limits but the tracing approach works.
AI-written summary · about
chapters (29)
- 0:00Introduction
- 2:19Agenda
- 3:57Transformation of infrastructure
- 7:59Monitoring
- 10:20Understanding alerts
- 13:13Building a picture
- 17:11Monitoring Kubernetes
- 18:30Proteus
- 18:45Kubernetes
- 19:42Cluster
- 20:13Service Master
- 20:34LK
- 21:15Different tools
- 21:55Service computing
- 25:38new challenges
- 27:04monitoring in a surrealist world
- 27:16option 1 outofthebox
- 27:47CloudWatch
- 29:33Creating your own solution
- 30:36Looking at logs
- 30:50Monitoring tools
- 33:18Demo
- 35:29What we need to remember
- 37:21Takeaways
- 38:02Loomigurumi
- 38:28Free tools
- 38:54Lambda cost
- 39:46Free trial
- 40:00Questions
related talks
24mObservability across serverless asynchronous managed services
35mMonitoring Microservices
39mBuilding Highly Scalable Microservices with Serverless
28mMonitoring Microservices w. Prometheus & Tracing Apps w. OpenTracing
11mLightning Talk: Searching for Serverless Candidates in the Jungle of Microservices
01h 00mMicroservices and Serverless
42mDistributed Monitoring: How to understand the Chaos
41mYou Don’t Want to Build a Serverless Monolith