KUBERNETES MONITORING

Kubernetes is a popular open source container management system and there has been growing customer requests for monitoring it inside AppDynamics.

Team

Product Manager – Mark Prichard
Engineering Lead – Steven Bolton
Sales Engineer – Sasha Jeltuhin

Problem

The current UI provided very minimal to no ability to monitor Kubernetes and this issue became one of the top sales blockers.

Storyboard

1. User discovers a pod with issues using a third party application.

2. The user uses multiple third-party applications to gather more event details.

3. The user opens AppDynamics and searches for the pod ID.

4. He guesses and checks every tier and node to find that particular pod ID. 

5. However, if the container got crushed already and no data would return.

6. The user needs to find out manual ways to solve the issue. 

Problems

1. The user needed to play a lot of guessing games to figure out where the pod with problems 

2. The user uses multiple third-party tools to gather Kubernetes related data.

Objective 1 - The Navigation

The most critical issue that needed to solve was the navigation. The proper navigation had to be provided to prevent repeated guesses. 

Objective 2 - Kubernetes Overview Dashboard

In parallel to initiating this project, Sasha was working on the Kubernetes overview dashboard. The 2nd goal was to integrate it with AppDynamics.

Objective 3 - Meaningful Data

The 3rd goal was to present just meaningful data/information for the user to reduce the frictions.

Proposed Flow

The proposed flow was created based on the design objectives.

Idea 1 - Using AppDynamics IA

The initial idea was created by using the current AppDynamics IA.

Idea 1 - The Challenge

1. AppDynamics IA was very different than how Kubernetes was structured 

2. The number of tabs could not get reduced down because of the technical limitations.

3. Kubernetes and AppDynamics had a discrepancy in using terminology (e.g. Node vs Node) 

Idea 2 - View AppDynamics through Kubernetes Lens

The schema selector enabled users to look into the current APM environment using the Kubernetes lens. WIthout the complete rework for the IA, this design provided more isolated view for Kubernetes. 

User Feedback - Mixed Feelings

"liked the idea that we're trying to show k8s design and flow: really interesting. Also liked that we're trying to build in a cluster view as part of product, surface HRs and events in k8s context."

"The flow doesn't feel right"

"We're talking k8s vs AppD nodes...Do we rename AppD nodes?

Final Decision - Top Level Navigation for Kubernetes Clusters

Rework on IA and provide the dedicated Kubernetes space. 

RESULT

Got all stakeholders’ and upper management's buy-in to rework on the IA and the project is moving towards the direction.

The backend work for cluster agents has begun to enable Kubernetes monitoring within AppDynamics.

Final Decision - Top Level Navigation for Kubernetes Clusters

Rework on IA and provide the dedicated space which also navigates to APM.

©2019 by Saeam Lee.