Introducing Log Tracking Feature to Revolutionize Kubernetes Observability

Explore a groundbreaking log tracking feature that transforms observability in Kubernetes in this 44-minute Linux Foundation conference talk. Delve into the fundamentals of Kubernetes (K8s) architecture, including declarative operations, cluster components, and the roles of Control Plane and Nodes. Discover the challenges in cross-component log investigation and learn about an innovative solution for tracking operation causes. Examine the implementation issues faced when dealing with multiple requests, chained component operations, and complex object processing scenarios. Gain insights into the current status of the proposed Kubernetes Enhancement Proposal (KEP) aimed at enhancing logging functionality and addressing these implementation challenges.

Intro
What is Kubernetes(k8s)?
Declarative operation Define the elements/conditions in yaml/json The definition describes the final state, not the procedure Can check the change history by version control such as Git All you have to do is change the code
What does k8s cluster consist of? The cluster consists of Control Plane and Nodes Control Plane Apiserver Controller manager
What is Control Plane doing? Make global decisions about the cluster, as well as detecting and responding to cluster events.
What is Node doing? Run the application deployed in the cluster Maintaining running pods and providing the Kubernetes runtime environment
Overview of K8s logging
The problem want to solve Difficult to investigate logs across components
Solution idea
Tracking the cause of operation Fujitsu
Check the impact of system failure
Overview of realization
Issues in implementation There are implementation issues in the following situations: Issue 1: Multiple requests are the cause of processing Issue 2: Multiple components operate in a chain with one request Issue 3: The object to be processed is different from the object of the Issue 4: The requesting and processing components are not one-to-one
Multiple requests are the cause of processing When multiple requests are processed at the same time, it is necessary to associate multiple IDs with the processing.
The object to be processed is different from the object of the trigger Ex. Pod update processes replicaset with trigger
Current status Proposing KEP to enhance the logging function. We are thinking of an implementation and system design that can solve all the implementation issues.