Meetup 22 - 04 - Logging and Monitoring at scale on Kubernetes
0 likes301 views
This document discusses logging and monitoring at scale on Kubernetes. It presents options for log handling in Kubernetes including files, transport via network protocols, and the console. It recommends the console for its broad support. It also discusses monitoring architecture with Operators to manage applications similarly to Kubernetes resources and provide lifecycle management. Finally, it proposes unifying metrics, logs, and traces user interfaces as current solutions have limitations and developers must check multiple places for different types of data.
Meetup 22 - 04 - Logging and Monitoring at scale on Kubernetes
- 1. Logging and Monitoring at scale on Kubernetes @dungdm93
- 2. 1. Logging 2
- 3. 1.1 Log Hander Option 1: File ● Pros ○ Easy to configs ○ Support by most services & libs ● Cons ○ Hard to Aggregate ○ No out-of-the box supported by k8s tools (e.g kubectl) 3
- 4. Option 2: Transport via Network (TCP/UDP/Socket) ● Pros ○ Easy to Aggregate ● Cons ○ Not every services & libs are supported ○ Each services/languages has different configuration ○ No out-of-the box supported by k8s tools (e.g kubectl) 1.1 Log Hander 4
- 5. Option 3: Console ● Pros ○ Supported by every services/languages ○ All official images logs to console by default ○ kubectl & other tools supported out-of-the-box ● Cons ○ Log format? 1.1 Log Hander 5
- 8. What is Operator? ● Custom Resource 8
- 9. What is Operator? ● Custom Resource ● Custom Resource Definition ● Controller 9
- 10. Benefit of Operator ● Cloud Native Approach ○ Manage App’s Resources in the same way as Kubernetes’ Resources. ● Lifecycle Management 10
- 12. 2.1 Monitoring Architecture (cont.) 12
- 13. 2.2 Monitor your own apps 13
- 14. 3. Recap 14 ● Logging ● Monitoring
- 16. 16
- 17. 17 Unify Metrics, Logs (and Traces) UI ElasticSearch & Kibana is good, but it has some limitations: ● ElasticSearch is far from what we need ● Kibana is slow ● Developer need to remember 2+ places: ○ 1 for Metrics ○ 1 for Logs ○ and… 1 more for Traces => Alternative solutions ● Loki + Grafana (evaluating…)
- 18. Thank you :) 18