AWS Builder Center

Builder Center

Troubleshooting a Multicloud CDN with AWS DevOps Agent.

Troubleshooting a Multicloud CDN with AWS DevOps Agent.

I deployed the new friontier AI agent tool, AWS DevOps Agent to troubleshoot and autonomously investigate incidents in a Multicloud Content Delivery Network running on AWS and Microsoft Azure.

Series: MultiCloud (1 article)

  1. 1
    Troubleshooting a Multicloud CDN with AWS DevOps Agent. This article

Ansley Adrian Moyo

Multi-cloud Advocate | Network Engineer | AWS | Microsoft Azure | Google Cloud | Cisco | Aviatrix
Being an Advocate for MultiCloud computing, demonstrated by previous projects and articles  I have writted on this Subject Matter, I was excited when AWS announced a new AI friontier agent, AWS DevOps Agent — an always-available teammate that spans software change and operations across AWS, multicloud, and on-prem environments, per official documentation .
The AWS DevOps Agent delivers its greatest value Post Deployment, by developing an understanding of your environment (Single or multicloud) along with your services, their dependencies, and production behavior. It autonomously investigates incidents, performs root cause analysis, recommends mitigation actions, and provides targeted insights to help prevent recurring issues and improve operational excellence and resilience. To demonstrate this function, I with my trusted AI Partner (AWS Kiro) teamed up and designed and deployed a MultiCloud CDN with an active origin in S3 and a failover origin in Microsoft Azure Blob storage.

MultiCloud Content Delivery Network ( AWS and Microsoft Azure)

Architecture Diagram

Technology Stack

  • OS Platform: Linux Ubuntu on WSL
  • AI Development: Kiro CLI (AI Agent) — IaC code generation, architecture design, and troubleshooting
  • Infrastructure as Code: Terraform — Multicloud resource provisioning
  • CDN: Amazon CloudFront — Global content delivery with edge caching
  • Primary Origin: Amazon S3 — Object storage for static assets
  • Failover Origin: Microsoft Azure Blob Storage (GRS) — Cross-cloud redundancy
  • Monitoring: Amazon CloudWatch — Metrics, alarms, and observability
  • Notifications: Amazon SNS — Alert routing to email/integrations
  • AI Operations: AWS DevOps Agent — Autonomous incident investigation and remediation
  • Identity Federation: AWS IAM + Azure Entra ID (OIDC) — Cross-cloud authentication for DevOps Agent
  • CLI Tools: AWS CLI, Azure CLI, GitHub CLI — Deployment and automation
  • Scripting: Bash — Chaos testing and DR simulation

Workflow Summary

Using the Kiro generated scripts and workflows, stored onto a local directory(and available on the github repo ), I started by provisioning the infrastructure with Terraform on Microsoft Azure and AWS.
On Microsoft Azure, I deployed the Blob Storage that serves as the failover origin
1
2
3
cd terraform/azure
terraform init
terraform apply
On AWS I provisioned a CloudFront distribution with an origin failover group, a Primary Origin(an S3 Bucket with Origin Access Control), CloudWatch Alarms(5xx, TotalErrorRate, latency, cache hit rate) and an SNS topic for email notifications.
1
2
3
4
5
cd terraform/aws
terraform init
terraform apply \
-var="azure_blob_endpoint=<BLOB_ENDPOINT_FROM_STEP_1>" \
-var="alert_email=your@email.com"

AWS DevOps Agent Setup

A critical step of this project is setting up the DevOps Agent, detailed comprehensively on Step 4 of the README .
Upon setting up the Agent Space, I moved on to registering the Azure Cloud resource, to enable the Agent to discover, investigate and remediate issues on Azure.
Registering the Azure Cloud with the agent, to achieve Cross-Cloud troubleshooting capabilities.
To interact with the DevOps Agent via a web interface, I navigated to the Web App tab within the Agent Space, enabled it, and clicked the Operator access link. From there I could interact with the agent in real time using the chat option, start investigations and view findings across both my AWS and Azure environments.
the Topology tab shows all your discovered resources in your multicloud environment
Through the Web App interface, you can interact with the Agent’s LLM, and ask questions using natural language prompts.
I asked the Agent to discover the resources in my multicloud environment

Chaos Testing

To fully test the capabilities of the Agent, there is need to engineer failure on one or both of the CDN’s origins. I asked Kiro to generate testing scripts, also available on the repo, to simulate DR Failover  and Total Outage  of the CDN.
a shell curl test, ran before the chaos test, to verify that traffic is being served from the Primary Origin

Test DR Failover

1
2
3
4
./simulate-failover.sh test # Verify primary is serving
./simulate-failover.sh break # Block S3, verify Azure takes over
./simulate-failover.sh restore # Restore S3
./simulate-failover.sh full # Run complete cycle
Failover Test success, showing the Blob Origin now serving traffic.
I then asked the Agent to investigate the issue, and suggest mitigations to solve this issue.
a Summary of the DevOps Agent findings, and the mitigation it suggested.

Test Total Outage

1
2
3
./simulate-total-outage.sh break-all # Block both origins
./simulate-total-outage.sh status # Check alarm states
./simulate-total-outage.sh restore-all # Restore everything
Kiro describing how the Total Outage chaos test is achieved.
once the script was ran, CloudWatch ran the 5xx error-rate alarm.
The Agent once again went to work to investigate the root cause.
Results of the Total Outage Investigation.
Once all the mitigations were applied, the CDN was restored back to serving traffic from the Primary Origin, as below.

Key Outcomes

Through designing, deploying and chaos testing, the following were the outcomes of the project
  • Autonomous incident resolution — AWS DevOps Agent reduced investigation time from hours of manual troubleshooting to minutes of automated root cause analysis across both clouds.
  • Multicloud resilience — Origin failover between AWS S3 and Azure Blob Storage delivered zero-downtime content delivery during simulated CSP failures.
  • Agentic AI in production operations — Demonstrated how frontier AI agents can continuously learn your environment, understand service dependencies, and act without human intervention when incidents occur.
  • Infrastructure as Code at scale — Terraform enabled repeatable, version-controlled deployment of multicloud infrastructure with independent AWS and Azure modules for flexible operations.
The entire repository is available on my Github, for future reference and replication — https://github.com/ansleyAA/multicloud-cdn-devops-agent 
# agentic-ai# amazon-cloudfront# multi-cloud# kiro# devops

Series: MultiCloud (1 article)

  1. 1
    Troubleshooting a Multicloud CDN with AWS DevOps Agent. This article
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article