<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by David Nithil on Medium]]></title>
        <description><![CDATA[Stories by David Nithil on Medium]]></description>
        <link>https://medium.com/@davidnithil?source=rss-e759794b975e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*sVAT6lKZ_kewaZ5Z.</url>
            <title>Stories by David Nithil on Medium</title>
            <link>https://medium.com/@davidnithil?source=rss-e759794b975e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 03:20:14 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@davidnithil/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Deploying a 3-Tier Microservices E-Commerce App on AKS using Helm, Azure Application Gateway…]]></title>
            <link>https://medium.com/@davidnithil/deploying-a-3-tier-microservices-e-commerce-app-on-aks-using-helm-azure-application-gateway-a19b057f5413?source=rss-e759794b975e------2</link>
            <guid isPermaLink="false">https://medium.com/p/a19b057f5413</guid>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[azure-kubernetes-service]]></category>
            <category><![CDATA[helm]]></category>
            <category><![CDATA[azure-application-gateway]]></category>
            <category><![CDATA[prometheus]]></category>
            <dc:creator><![CDATA[David Nithil]]></dc:creator>
            <pubDate>Thu, 09 Apr 2026 18:12:51 GMT</pubDate>
            <atom:updated>2026-04-09T18:12:51.126Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Deploying a 3-Tier Microservices E-Commerce App on AKS using Helm, Azure Application Gateway, Prometheus &amp; Grafana</strong></h3><h3>Introduction</h3><p>Modern cloud-native applications are no longer built as one large monolith. Instead, they are split into smaller, independently deployable services that can scale, recover, and evolve faster. To gain hands-on experience with this architecture, I deployed <strong>Stan’s Robot Shop</strong>, a <strong>microservices-based e-commerce demo application</strong>, on <strong>Azure Kubernetes Service (AKS)</strong> using <strong>Helm</strong>.</p><p>This project helped me simulate a realistic DevOps and Kubernetes deployment flow with:</p><ul><li><strong>Azure Kubernetes Service (AKS)</strong></li><li><strong>Helm charts for application deployment</strong></li><li><strong>Azure Application Gateway Ingress Controller (AGIC)</strong></li><li><strong>Persistent Volumes / Persistent Volume Claims</strong></li><li><strong>Prometheus-based monitoring</strong></li><li><strong>Grafana dashboards</strong></li><li><strong>Oracle Cloud Free Tier Ubuntu VM</strong> as an admin/bastion machine to access and manage the AKS cluster using kubectl</li></ul><p>The goal was not just to “make the app work,” but to build and validate a deployment that reflects how real-world cloud-native applications are operated.</p><h3>Project Overview</h3><p>The application used in this project is <strong>Stan’s Robot Shop</strong>, a sample <strong>microservices e-commerce platform</strong> containing <strong>11 services</strong> working together to serve a web storefront.</p><p>The application includes services such as:</p><ul><li><strong>web</strong></li><li><strong>user</strong></li><li><strong>cart</strong></li><li><strong>catalogue</strong></li><li><strong>shipping</strong></li><li><strong>payment</strong></li><li><strong>dispatch</strong></li><li><strong>ratings</strong></li><li><strong>mongodb</strong></li><li><strong>mysql</strong></li><li><strong>redis</strong></li><li>plus supporting messaging and infrastructure components</li></ul><p>This makes it an excellent project to demonstrate:</p><ul><li>Kubernetes workload management</li><li>Service discovery</li><li>Ingress routing</li><li>Stateful and stateless application handling</li><li>Monitoring and observability</li><li>Helm-based deployments</li></ul><h3>Architecture Used</h3><h3>High-Level Flow</h3><p>The architecture for this demo looked like this:</p><ol><li>I used an <strong>Oracle Cloud Ubuntu VM</strong> as my administration machine.</li><li>From that VM, I installed and configured:</li></ol><ul><li>kubectl</li><li>helm</li><li>Azure CLI</li></ul><p>3. I created and managed an <strong>AKS cluster in Azure</strong>.</p><p>4. The Robot Shop application was deployed into AKS using <strong>Helm</strong>.</p><p>5. External traffic was routed through <strong>Azure Application Gateway Ingress</strong>.</p><p>6. Monitoring was enabled using:</p><ul><li><strong>Azure Managed Prometheus</strong></li><li><strong>Grafana dashboards</strong></li></ul><p>7. Persistent application components used <strong>Kubernetes PVCs</strong> for storage.</p><p>This setup reflects a common enterprise pattern where engineers manage cloud infrastructure from a secure Linux admin host or jump box.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MlKxxEvnyDNaEJvRCJUWwQ.png" /><figcaption>Blueprint of a Cloud-Native Ecommerce Stack (Neon Edition)</figcaption></figure><h3>Why I Used an Oracle Cloud Free Tier Ubuntu VM</h3><p>One interesting aspect of this project is that I used an <strong>Oracle Cloud Free Tier Ubuntu VM</strong> to manage my <strong>Azure Kubernetes cluster</strong>.</p><p>This approach is useful because it simulates a real-world operations model where:</p><ul><li>Kubernetes clusters are managed from a <strong>dedicated Linux administration machine</strong></li><li>Tooling is centralized</li><li>Access and deployment workflows are consistent</li><li>Engineers can manage multiple cloud environments from one operational host</li></ul><p>On this VM, I used:</p><ul><li>kubectl for cluster operations</li><li>helm for package deployment</li><li>shell commands for troubleshooting and validation</li></ul><p>This also gave me practical cross-cloud operational experience.</p><h3>Azure Resources Created</h3><p>After provisioning AKS and related services, Azure created several supporting resources automatically and manually, including:</p><ul><li><strong>AKS cluster</strong></li><li><strong>Virtual Network</strong></li><li><strong>VM Scale Set</strong> for worker nodes</li><li><strong>Network Security Group</strong></li><li><strong>Application Gateway</strong></li><li><strong>Public IP</strong></li><li><strong>Azure Monitor workspace</strong></li><li><strong>Managed Grafana</strong></li><li><strong>Managed Identity</strong></li><li>supporting monitoring and diagnostic resources</li></ul><p>This is a valuable learning point for anyone working with AKS:<br> creating a Kubernetes cluster in Azure is not just “one resource” — it provisions and integrates multiple Azure services behind the scenes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a1JnfX8MMkQkPa6Sk5VuHw.png" /><figcaption>(Shows all AKS-related Azure resources)</figcaption></figure><h3>AKS Cluster Setup</h3><p>I provisioned an <strong>AKS cluster</strong> and validated the node readiness from my Ubuntu VM.</p><p>The cluster had <strong>2 worker nodes</strong>, which is sufficient for a demo microservices environment and helps simulate workload scheduling and service distribution across multiple nodes.</p><h3>Node Validation</h3><p>I validated the nodes using:</p><blockquote>kubectl get nodes</blockquote><p>This confirmed both worker nodes were in the <strong>Ready</strong> state.</p><p>This is a critical validation step because even if a cluster is provisioned successfully, workloads cannot be scheduled unless the nodes are healthy and available.</p><h3>Deploying the Application with Helm</h3><p>Instead of manually creating every Kubernetes object using raw YAML, I deployed the application using <strong>Helm</strong>.</p><h3>Why Helm?</h3><p>Helm is often referred to as the <strong>package manager for Kubernetes</strong>. It helps in:</p><ul><li>Packaging complex applications</li><li>Managing reusable templates</li><li>Standardizing deployments</li><li>Simplifying upgrades and rollbacks</li><li>Reducing repetitive YAML management</li></ul><p>For a multi-service application like Robot Shop, Helm is the right choice because it organizes deployment manifests cleanly and makes the release process much more manageable.</p><h3>Deployment Scope</h3><p>Using Helm, I deployed:</p><ul><li>Deployments / ReplicaSets</li><li>Services</li><li>Ingress resources</li><li>PVC-backed components</li><li>namespace-scoped workloads</li></ul><p>This made the deployment structured, repeatable, and closer to real-world production practices.</p><h3>Namespace and Workload Isolation</h3><p>I deployed the application into a dedicated namespace: robot-shop</p><p>This is a best practice because it helps isolate application resources from system workloads like:</p><ul><li>kube-system</li><li>monitoring components</li><li>cluster add-ons</li></ul><p>It also makes troubleshooting and operations easier.</p><blockquote>kubectl get all -n robot-shop<br>kubectl get svc -n robot-shop<br>kubectl get pods -n robot-shop</blockquote><p>Using namespaces is a small thing, but it reflects mature Kubernetes operational habits.</p><h3>Microservices Running in AKS</h3><p>One of the most satisfying parts of this project was seeing all the microservices deployed and communicating successfully inside AKS.</p><p>From the Kubernetes output, I verified the application services such as:</p><ul><li>cart</li><li>catalogue</li><li>dispatch</li><li>mongodb</li><li>mysql</li><li>payment</li><li>rabbitmq</li><li>ratings</li><li>redis</li><li>shipping</li><li>user</li><li>web</li></ul><p>Each service was exposed internally via <strong>ClusterIP</strong>, while the frontend web service was exposed externally through <strong>LoadBalancer/Ingress pathing</strong> depending on the traffic flow.</p><p>This demonstrates how Kubernetes handles:</p><ul><li>internal service discovery</li><li>inter-service communication</li><li>frontend/backend separation</li><li>stateful and stateless workload coexistence</li></ul><h3>Persistent Storage Validation</h3><p>Stateful services like databases and caches often need persistence.</p><p>In this deployment, I validated the presence of <strong>Persistent Volume Claims (PVCs)</strong>, including a bound PVC for Redis:</p><pre>kubectl get pvc -n robot-shop</pre><p>The PVC was in <strong>Bound</strong> state, confirming that Kubernetes successfully attached persistent storage(I used azure disks) to the workload.</p><h3>Why This Matters</h3><p>Without persistent storage:</p><ul><li>databases lose data after pod restarts</li><li>cache layers reset unexpectedly</li><li>application reliability suffers</li></ul><p>Persistence is one of the key differences between a toy Kubernetes deployment and a more realistic one.</p><h3>Ingress Configuration with Azure Application Gateway</h3><p>One of the most important parts of this deployment was exposing the application externally using <strong>Azure Application Gateway Ingress</strong>.</p><h3>Why I Used Azure Application Gateway</h3><p>Instead of relying only on a standard LoadBalancer service, Azure Application Gateway provides a more enterprise-grade ingress layer with benefits such as:</p><ul><li>Layer 7 routing</li><li>better ingress management</li><li>Azure-native integration</li><li>centralized entry point for application traffic</li></ul><p>I validated the ingress resource using:</p><pre>kubectl get ing -n robot-shop<br>kubectl logs deploy/ingress-appgw-deployment -n kube-system</pre><p>This showed that the Robot Shop application was mapped to a public IP.</p><p>The ingress was associated with:</p><ul><li>an <strong>Azure Application Gateway</strong></li><li>a <strong>public IP address</strong></li><li>HTTP routing to the application frontend</li></ul><p>This is closer to how production workloads are exposed in enterprise environments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g6v_yun7PvWYe27qce9Isw.png" /><figcaption>(Shows nodes, services, replicasets, PVC, ingress)</figcaption></figure><h3>Accessing the Application</h3><p>Once the ingress and services were configured correctly, I was able to access the application in a browser through the public IP.</p><p>The <strong>Stan’s Robot Shop</strong> UI loaded successfully, confirming that:</p><ul><li>frontend traffic was reaching the cluster</li><li>backend microservices were responding correctly</li><li>service-to-service communication was healthy</li><li>ingress routing was working as expected</li></ul><p>Seeing the full UI load is important because it proves that the deployment is not just “Kubernetes objects created successfully” — it confirms <strong>end-to-end application functionality</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eiiMR5KP1N9Pxil2Vpw3Qg.png" /><figcaption>Stan’s Robot Shop live ecommerce website</figcaption></figure><h3>Monitoring the AKS Cluster</h3><p>Deploying an application is only one part of DevOps.<br> The next important step is <strong>observability</strong>.</p><p>To monitor the health and performance of the AKS cluster and workloads, I enabled:</p><ul><li><strong>Prometheus-based metrics</strong></li><li><strong>Grafana dashboards</strong></li><li><strong>Azure Monitor integration</strong></li></ul><p>This gave me visibility into:</p><ul><li>node usage</li><li>pod count</li><li>namespace resource consumption</li><li>memory trends</li><li>CPU trends</li><li>disk activity</li><li>cluster health indicators</li></ul><h3>Prometheus Monitoring</h3><p>From the AKS <strong>Monitor</strong> view, I was able to observe the cluster through <strong>Managed Prometheus visualizations</strong>.</p><p>The monitoring page showed:</p><ul><li><strong>2 total nodes</strong></li><li><strong>44 total pods</strong></li><li>namespace-level CPU consumption</li><li>cluster overview and health signals</li></ul><p>This is useful because it allows engineers to quickly answer questions like:</p><ul><li>Is the cluster healthy?</li><li>Are workloads running as expected?</li><li>Are any namespaces consuming unusual CPU?</li><li>Are pods scaling or failing unexpectedly?</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*trcV7nuYmMKVKTzZWOQV-g.png" /><figcaption>(Managed Prometheus visualization)</figcaption></figure><h3>Grafana Dashboards</h3><p>I also enabled <strong>Grafana dashboards</strong> to visualize system metrics in a more operational and dashboard-friendly way.</p><p>The Grafana view showed useful metrics such as:</p><ul><li><strong>CPU usage percentage</strong></li><li><strong>load average</strong></li><li><strong>memory usage</strong></li><li><strong>memory free / cached / buffers</strong></li><li>node-level observability</li></ul><p>One of the visible metrics from the dashboard was <strong>memory usage around 30%</strong>, which indicates the cluster was running comfortably under load during the observed period.</p><h3>Why Grafana Matters?</h3><p>Grafana is extremely useful in real-world operations because it helps teams:</p><ul><li>visualize trends quickly</li><li>correlate issues faster</li><li>identify resource bottlenecks</li><li>improve troubleshooting efficiency</li></ul><p>Monitoring is what turns a deployment into an operable platform.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TIhuWrHKd5QXtVexPXWmeQ.png" /><figcaption>(Node explorer — one of the default dashboard)</figcaption></figure><h3>What I Learned from This Project?</h3><p>This project was much more than just deploying a demo app.</p><p>It gave me hands-on practice in multiple DevOps and Kubernetes areas:</p><h3>1. Helm-Based Application Management</h3><p>I got practical experience in deploying and managing a complex microservices app using Helm instead of manual YAML sprawl.</p><h3>2. AKS Cluster Operations</h3><p>I worked with real AKS resources, nodes, networking, and validation workflows.</p><h3>3. Ingress and Traffic Routing</h3><p>Configuring and validating <strong>Azure Application Gateway Ingress</strong> gave me experience with production-style traffic exposure.</p><h3>4. Stateful Workloads in Kubernetes</h3><p>By validating PVC-backed services, I reinforced how persistence works in Kubernetes.</p><h3>5. Monitoring and Observability</h3><p>Prometheus and Grafana helped me understand how to observe cluster health and workload behavior after deployment.</p><h3>6. End-to-End Troubleshooting Mindset</h3><p>Projects like this naturally require checking:</p><ul><li>pod readiness</li><li>service connectivity</li><li>ingress mapping</li><li>PVC state</li><li>namespace isolation</li><li>external accessibility</li><li>cluster health metrics</li></ul><p>That operational troubleshooting mindset is exactly what makes DevOps work valuable beyond just writing manifests.</p><h3>Why This Project Matters for DevOps Engineers?</h3><p>A lot of people stop after saying:</p><p><em>“I deployed an app on Kubernetes.”</em></p><p>But what really matters is whether you can also answer:</p><ul><li>How is traffic entering the cluster?</li><li>How are services communicating internally?</li><li>How is persistence handled?</li><li>How do you monitor workloads?</li><li>How do you validate the cluster after deployment?</li><li>How do you troubleshoot issues if the app is not accessible?</li></ul><p>This project helped me connect those pieces together into one working implementation.</p><p>That is what transforms a simple Kubernetes demo into a <strong>real DevOps showcase project</strong>.</p><h3>Next Improvements I’d Add (Not immediately😜 )</h3><p>If I continue evolving this project, the next logical improvements would be:</p><ul><li><strong>CI/CD pipeline integration — </strong>GitHub Actions / Azure DevOps pipelines</li><li><strong>Argo CD / GitOps deployment model</strong></li><li><strong>TLS / HTTPS enablement</strong></li><li><strong>custom domain mapping</strong></li><li><strong>Horizontal Pod Autoscaler (HPA)</strong></li><li><strong>Pod disruption budgets</strong></li><li><strong>resource requests and limits tuning</strong></li><li><strong>centralized logging</strong></li><li><strong>security hardening</strong></li><li>RBAC</li><li>secrets management</li><li>image scanning</li><li>policy enforcement</li></ul><p>These would make the deployment even closer to a production-ready platform.</p><h3>Conclusion</h3><p>This project was a great hands-on exercise in deploying and operating a <strong>microservices-based 3-tier e-commerce application</strong> on <strong>Azure Kubernetes Service</strong> using <strong>Helm</strong>, <strong>Azure Application Gateway Ingress</strong>, <strong>Prometheus</strong>, and <strong>Grafana</strong>.</p><p>It gave me practical experience across:</p><ul><li>Kubernetes deployment</li><li>Helm packaging</li><li>ingress exposure</li><li>persistent storage</li><li>observability</li><li>cluster validation</li><li>real-world DevOps troubleshooting</li></ul><p>More importantly, it reinforced that successful cloud-native deployments are not just about “running containers,” but about building systems that are <strong>deployable, observable, and operable</strong>.</p><p>If you’re learning Kubernetes, AKS, or DevOps, this kind of project is one of the best ways to move from theory into real implementation🐱‍🏍.</p><p><strong>GitHub Repo Used:</strong><br> <a href="https://github.com/davidnithil/three-tier-architecture-demo">https://github.com/davidnithil/three-tier-architecture-demo</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a19b057f5413" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>