Use Gemini Cloud Assist to investigate and troubleshoot service issues in Google Cloud

Key Concepts

• Observability: The ability to understand the internal state of a system from its external outputs.
• Troubleshooting: The process of diagnosing and resolving problems in a system.
• Unknown Unknowns: Problems that are not anticipated or understood.
• Gemini Cloud Assist: A Google Cloud tool designed to aid in troubleshooting.
• Investigations: A feature within Gemini Cloud Assist that analyzes symptoms, identifies underlying issues, and recommends solutions.
• Google Kubernetes Engine (GKE): A managed Kubernetes service provided by Google Cloud.
• CrashLoopBackOff: A Kubernetes error state where a container repeatedly crashes and restarts.
• Cloud Logging: Google Cloud's centralized logging service.
• Generative AI: Artificial intelligence capable of generating new content, such as text or code.

Investigations Tool in Gemini Cloud Assist: A Troubleshooting Aid

Introduction

Kyle Benson, a product manager at Google Cloud, introduces the Investigations tool within Gemini Cloud Assist, designed to simplify troubleshooting in Google Cloud environments. The tool aims to address the challenges posed by "unknown unknowns" by providing data and tools for critical problem-solving.

Functionality of the Investigations Feature

The Investigations feature assists in the troubleshooting process by:

• Taking symptoms or error messages as input.
• Identifying underlying issues.
• Recommending actionable solutions.

It can be initiated from scratch or within the context of various Google Cloud services.

Example: Troubleshooting a GKE Application Issue

1. Problem Identification: A notification indicates an issue with a service running in Google Kubernetes Engine (GKE). Specifically, the "inventory app" is experiencing a "CrashLoopBackOff" error and lacks minimum availability.
2. Log Analysis: The engineer accesses container logs for the service using Cloud Logging to investigate the cause of the crashes. Intermittent errors are observed.
3. Initiating an Investigation: Upon clicking on a recent error in Cloud Logging, the "Investigate" button appears, allowing the engineer to create a new investigation.
4. Configuring the Investigation:
   • The investigation is named "System is down!!" to indicate urgency.
   • The start time is left at the default.
   • The issue type is identified as "I have an error message," leveraging the prepopulated error information.
   • Relevant resources are kept at their default settings.
5. Analysis and Findings: Gemini analyzes the relevant resources and identifies a potential error on line 14 within the Python code of the application.
6. Interpretation and Action: The engineer interprets this finding as a recent code change introducing an error that causes the container to crash. The recommended action is to:
   • Contact the team responsible for the application.
   • Reroll the problematic package.
   • Reroll the deployment.

Further Support and Integration

• If the issue persists after trying the suggested solutions, users can create a case with Google Cloud Support, providing the steps already attempted with Gemini.
• Investigations can be initiated from various points within Google Cloud, including Logs Explorer, Cloud Console chat, and other services to be introduced.

Benefits and Conclusion

The Investigations feature aims to redefine troubleshooting by:

• Leveraging generative AI to surface insights instantly.
• Guiding decision-making.
• Helping resolve issues faster.

Users are encouraged to try Gemini Cloud Assist and provide feedback.