<?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 Kyle Law on Medium]]></title>
        <description><![CDATA[Stories by Kyle Law on Medium]]></description>
        <link>https://medium.com/@kylelzk?source=rss-e060f219712c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*pV5bgZL59kGmgzHAzmcsDw.png</url>
            <title>Stories by Kyle Law on Medium</title>
            <link>https://medium.com/@kylelzk?source=rss-e060f219712c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 15:31:25 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kylelzk/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[[Lab] Incident Response with AI Agent in Kubernetes]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@kylelzk/lab-ai-agent-in-kubernetes-76af66018368?source=rss-e060f219712c------2"><img src="https://cdn-images-1.medium.com/max/1400/1*DCBTYOH-chEp0PX2Y0p4Ew.png" width="1400"></a></p><p class="medium-feed-snippet">I gave an AI agent read only access to my Kubernetes cluster. Here&#x2019;s what happened.</p><p class="medium-feed-link"><a href="https://medium.com/@kylelzk/lab-ai-agent-in-kubernetes-76af66018368?source=rss-e060f219712c------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@kylelzk/lab-ai-agent-in-kubernetes-76af66018368?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/76af66018368</guid>
            <category><![CDATA[incident-response]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[observability]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Sun, 18 Oct 2026 17:00:00 GMT</pubDate>
            <atom:updated>2026-10-18T17:32:59.118Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[[CKNE Exam Prep] 3.3 — Configuring Egress Gateways for Cluster Exit Traffic]]></title>
            <link>https://medium.com/@kylelzk/ckne-exam-prep-3-3-configuring-egress-gateways-for-cluster-exit-traffic-ad1675c33a61?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ad1675c33a61</guid>
            <category><![CDATA[cilium]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[egress-gateway]]></category>
            <category><![CDATA[ckne]]></category>
            <category><![CDATA[kubernetes-networking]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Sat, 26 Sep 2026 13:38:22 GMT</pubDate>
            <atom:updated>2026-09-26T13:38:22.746Z</atom:updated>
            <content:encoded><![CDATA[<p>Egress means traffic out of cluster, whereas Ingress is traffic into the cluster. In this article, I’ll be sharing how to setup Egress Gateway.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lgAmrvCsrBDVc8ic8_hobg.png" /><figcaption>Egress Gateway</figcaption></figure><p>Ways to setup: Cilium / Istio (or more…)</p><h3>How to setup with Cilium?</h3><ol><li><strong>Enable KP replacement &amp; BPF masquerading &amp; egressGateway</strong></li></ol><pre>helm upgrade cilium cilium/cilium -n kube-system --reuse-values \<br>  --set egressGateway.enabled=true \<br>  --set bpf.masquerade=true \<br>  --set kubeProxyReplacement=true<br><br>kubectl rollout restart ds/cilium -n kube-system<br>kubectl rollout restart deploy/cilium-operator -n kube-system</pre><p>So this part, it’ll install a CRD named CiliumEgressGatewayPolicies (cegp), which you’ll use in step 3</p><pre>k api-resources  | grep egress<br>ciliumegressgatewaypolicies         cegp                                cilium.io/v2                      false        CiliumEgressGatewayPolicy</pre><p><strong>2. Label node to be Egress node</strong></p><pre>k label node &lt;node-name&gt; egress-node=true</pre><p><strong>3. Create </strong><strong>CiliumEgressgatewayPolicy (cegp)</strong></p><pre>apiVersion: cilium.io/v2<br>kind: CiliumEgressGatewayPolicy<br>metadata:<br>  name: myapp-egress<br>spec:<br>  selectors:<br>  - podSelector:<br>      matchLabels:<br>        app: myapp<br>  destinationCIDRs:<br>  - &quot;0.0.0.0/0&quot;<br>  egressGateway:<br>    nodeSelector:<br>      matchLabels:<br>        egress-node: &quot;true&quot;<br>    egressIP: 10.168.60.100   # optional; must exist on the gateway node&#39;s interface</pre><p>cepg allows you to decide:</p><ul><li>decides which deployment</li><li>which destination</li><li>which IP interface on the node ( a node can have ≥1 IP address)</li><li>… more</li></ul><p>to route through the Egress node</p><p><strong>4. Test if traffic from pod with label </strong><strong>app: myapp go through the egress-node, eg.</strong></p><pre>kubectl run egress-test \<br>  --image=curlimages/curl:8.10.1 \<br>  --labels=app=myapp \<br>  --overrides=&#39;{&quot;spec&quot;:{&quot;nodeName&quot;:&quot;&lt;OTHER NODE THAN THE EGRESS NODE&gt;&quot;}}&#39; \<br>  --command -- sleep infinity</pre><p>How to validate:</p><ol><li>From the pod, curl ifconfig.me / api.ipify.org / icanhazip.com / checkip.amazonaws.com / httpbin.org/ip <br>All these links would return the egress IP</li><li>If your environment go through a upstream provider NAT, you could setup an internal server on a node that’s not on the cluster</li></ol><pre># On a node under the same Internal Network, but out of the cluster<br>python3 -m http.server 8080</pre><p>Then curl the internal endpoint from the cluster</p><pre>k exec deploy/app -- curl -s http://&lt;other-internal-node-ip&gt;:8080</pre><p>From the python server, you’ll see the Egress Node IP</p><pre>Serving HTTP on 0.0.0.0 port 8080 (http://0.0.0.0:8080/) ...<br>172.16.0.4 - - [26/Sep/2026 13:19:42] &quot;GET / HTTP/1.1&quot; 200 -<br>172.16.0.4 - - [26/Sep/2026 13:20:27] &quot;GET / HTTP/1.1&quot; 200 -<br>172.16.0.4 - - [26/Sep/2026 13:27:34] &quot;GET / HTTP/1.1&quot; 200 -</pre><p>Yep so, here is how you could cherry pick a node to be an Egress Node, and route a particular deployment / destination traffic through that node with Cilium Egress Gateway.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X1z_ySHexA6pVDmDWwzfZA.png" /><figcaption>Architecture</figcaption></figure><h4>Kyle, how about through Istio instead of Cilium?</h4><p>next time 😅</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ad1675c33a61" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[[Kubernetes Lab] Understanding Gateway API Inference Extension]]></title>
            <link>https://medium.com/@kylelzk/kubernetes-lab-understanding-gateway-api-inference-extension-68b948f2d721?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/68b948f2d721</guid>
            <category><![CDATA[ckne]]></category>
            <category><![CDATA[llm-inference]]></category>
            <category><![CDATA[kubernetes-networking]]></category>
            <category><![CDATA[inference]]></category>
            <category><![CDATA[kubernetes]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 15:08:59 GMT</pubDate>
            <atom:updated>2026-09-25T15:08:59.662Z</atom:updated>
            <content:encoded><![CDATA[<p>If you have an inference service running on Kubernetes, how could you determine which pods to send traffic to? This is where <strong>Inference Extension </strong>comes into play. This article would show you how to setup end-to-end North-South Networking for Inference on Kubernetes.</p><blockquote>This is a practical walkthrough guide, and expected you understand basic concepts of Gateway API.<br>We’ll be using vllm-d-inference-sim image, which is a simulator for inference and can be run on CPU-only Kubernetes cluster.</blockquote><p>Video walkthrough for this article: <a href="https://youtu.be/ozSN-_CWdUg">https://youtu.be/ozSN-_CWdUg</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RAS3t49RQ3clVmzKRoLyYQ.png" /></figure><h4>What is Gateway API Inference extension?</h4><p>Gateway API Inference extension is an extension from Gateway API for Inferences workload (ohh wow…)</p><p>Basically it contains 1 new CRD, InferencePool (there’s also InferenceObjective and InferencePoolImport CRDs we’re not going there in this lab) and EndpointPicker (EPP), which is a set of deploy-service k8s objects, using EPP docker image with gRPC.</p><h4>The Problem</h4><p>Unlike normal deployment, inference has to be scheduled on less utilized pod based on specific metrics. For the best inference performance, we need to route traffic to least utilized LLM pod based on some metrics.</p><p>This borned the extension.</p><ul><li>EPP basically continuously scrape LLM pod metrics and tell you which pod is less utilized</li><li>Inference Pool (<em>infpool</em>) connects 2 deployments together : LLM and EPP</li><li>Then <em>infpool</em> is referenced by HTTPRoute.spec.rules.backendRef</li></ul><h4>In other words…</h4><p>InferencePool is just another backendRefs for Gateway, i.e.</p><pre>Gateway -&gt; HTTPRoute -&gt; InferencePool (instead of Service) -&gt; EndpointPicker</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vecqymtZdJFECE8hPako5Q.png" /><figcaption>Gateway API Inference Extension</figcaption></figure><h3>The Setup</h3><h4>Stage 1 — The LLM</h4><ol><li>Deploy this <a href="https://github.com/llm-d/llm-d-inference-sim/blob/main/manifests/deployment.yaml">https://github.com/llm-d/llm-d-inference-sim/blob/main/manifests/deployment.yaml</a>, basically a deployment and service for simulated LLM</li><li>After deployed, run curl &lt;svc IP&gt;:&lt;svc port&gt;/v1/models , or simulated response with</li></ol><pre>curl http://&lt;svc IP&gt;:&lt;svc port&gt;/v1/chat/completions \<br>  -H &quot;Content-Type: application/json&quot; \<br>  -d &#39;{<br>    &quot;model&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>    &quot;messages&quot;: [<br>      {&quot;role&quot;: &quot;user&quot;, &quot;content&quot;: &quot;Explain Kubernetes in one sentence.&quot;}<br>    ],<br>    &quot;max_tokens&quot;: 256,<br>    &quot;temperature&quot;: 0.7<br>  }&#39;</pre><p>The response is simulated</p><blockquote>Tada! You got a simplest LLM inference running with 1 deploy and 1 svc.</blockquote><h4>Stage 2 — The Gateway API</h4><p>So now, instead of using the service cluster IP, I want to access it through a Gateway, with a domain name and a path, such like</p><p>curl kyle-is-not-handsome.com/inference/v1/chat/completions</p><p>To achieve this, the concept is like:</p><blockquote>We need a <strong>gateway</strong> that use domain: kyle-is-not-handsome.com , then a <strong>httproute</strong> adding /inference on that domain, mapping to the service.</blockquote><p>The step-by-step is like:</p><ol><li>Install Gateway API CRD</li><li>Install Gateway Class (eg. istio, nginx, etc…)</li><li>Setup Gateway refer to the gc in (1), mention the domain</li><li>Setup httproute, add /inference path, map to the service</li><li>(locally) add &lt;gateway IP&gt; &lt;domain&gt; into /etc/hosts, so the domain would be mapped into the gateway IP.</li></ol><blockquote>Let’s do it!</blockquote><ol><li>Install Gateway API CRD</li></ol><p><em>Before and After</em>: run k api-resources | grep gateway , you’ll see several gateway components like GatewayClass, Gateway, HTTPRoute after installation</p><pre>kubectl get crd gateways.gateway.networking.k8s.io &amp;&gt; /dev/null || \<br>  { kubectl kustomize &quot;github.com/kubernetes-sigs/gateway-api/config/crd?ref=v1.6.0&quot; | kubectl apply -f -; }</pre><p>2. Install GatewayClass</p><p>I use istio Gateway Class here:</p><ul><li>2.1) Install istioctl: <a href="https://istio.io/latest/docs/setup/install/istioctl/">https://istio.io/latest/docs/setup/install/istioctl/</a></li><li>2.2) istioctl install --set profile=minimal -y</li><li>2.3) Check GatewayClass</li></ul><pre>ubuntu@k8s-t7-cp1:~$ k get gc<br>NAME           CONTROLLER                    ACCEPTED   AGE<br>istio          istio.io/gateway-controller   True       68m<br>istio-remote   istio.io/unmanaged-gateway    True       68m</pre><p>3. Setup Gateway</p><pre>apiVersion: gateway.networking.k8s.io/v1<br>kind: Gateway<br>metadata:<br>  name: inf-gtw<br>  namespace: inf<br>  annotations:<br>    networking.istio.io/service-type: NodePort<br>spec:<br>  gatewayClassName: istio<br>  listeners:<br>  - name: http<br>    protocol: HTTP<br>    port: 80<br>    hostname: &quot;kyle-is-not-handsome.com&quot;</pre><p>Note: For local testing and istio gc, you need to add annotations to <em>NodePort</em>. By default it’s using Loadbalancer.</p><p>After apply, run k get svc to observe the corresponding NodePort svc created from the gtw</p><p>4. Setup HTTPRoute</p><pre>apiVersion: gateway.networking.k8s.io/v1<br>kind: HTTPRoute<br>metadata:<br>  name: inf-route<br>  namespace: inf<br>spec:<br>  parentRefs:<br>  - name: inf-gtw<br>  hostnames:<br>  - &quot;kyle-is-not-handsome.com&quot;<br>  rules:<br>  - matches:<br>    - path:<br>        type: PathPrefix<br>        value: /inference<br>    filters:<br>    - type: URLRewrite<br>      urlRewrite:<br>        path:<br>          type: ReplacePrefixMatch<br>          replacePrefixMatch: /<br>    backendRefs:<br>    - name: vllm-sim-demo-svc<br>      port: 8000</pre><p>5. add DNS mapping into /etc/hosts</p><pre>&lt;gtw svc IP&gt; kyle-is-not-handsome.com</pre><p>6. Check you result</p><pre>ubuntu@k8s-t7-cp1:~/inf$ curl kyle-is-not-handsome.com/inference/v1/chat/completions   -H &quot;Content-Type: application/json&quot;   -d &#39;{<br>    &quot;model&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>    &quot;messages&quot;: [<br>      {&quot;role&quot;: &quot;user&quot;, &quot;content&quot;: &quot;Explain Kubernetes in one sentence.&quot;}<br>    ],<br>    &quot;max_tokens&quot;: 256,<br>    &quot;temperature&quot;: 0.7<br>  }&#39; | jq<br>{<br>  &quot;id&quot;: &quot;chatcmpl-81f105fe-8a03-511c-a500-96351edb247a&quot;,<br>  &quot;model&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>  &quot;object&quot;: &quot;chat.completion&quot;,<br>  &quot;created&quot;: 1790150848,<br>  &quot;usage&quot;: {<br>    &quot;prompt_tokens&quot;: 17,<br>    &quot;completion_tokens&quot;: 39,<br>    &quot;total_tokens&quot;: 56,<br>    &quot;prompt_tokens_details&quot;: {<br>      &quot;cached_tokens&quot;: 0<br>    }<br>  },<br>  &quot;choices&quot;: [<br>    {<br>      &quot;index&quot;: 0,<br>      &quot;finish_reason&quot;: &quot;stop&quot;,<br>      &quot;message&quot;: {<br>        &quot;role&quot;: &quot;assistant&quot;,<br>        &quot;content&quot;: &quot;To be or not to be that is the question. Give a man a fish and you feed him for a day; teach a man to fish and you feed him for a lifetime Alas&quot;<br>      }<br>    }<br>  ]<br>}<br>ubuntu@k8s-t7-cp1:~/inf$ curl kyle-is-not-handsome.com/inference/v1/models  | jq<br>{<br>  &quot;object&quot;: &quot;list&quot;,<br>  &quot;data&quot;: [<br>    {<br>      &quot;id&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>      &quot;object&quot;: &quot;model&quot;,<br>      &quot;created&quot;: 1790150850,<br>      &quot;owned_by&quot;: &quot;vllm&quot;,<br>      &quot;root&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>      &quot;parent&quot;: null,<br>      &quot;max_model_len&quot;: 1024<br>    }<br>  ]<br>}</pre><blockquote>yeah, kyle-is-not-handsome.com/inference is now mapping to the LLM service</blockquote><p>So far, we’re just utilizing the NORMAL Gateway API usage, next we’re adding the Inference Extension</p><h4>Stage 3: The Idol of this article — — — — Inference Extension!</h4><ol><li>Install CRDs</li></ol><p>kubectl apply -k <a href="https://github.com/kubernetes-sigs/gateway-api-inference-extension/config/crd">https://github.com/kubernetes-sigs/gateway-api-inference-extension/config/<strong>crd</strong></a></p><p>1.1 Enable Istio extension (based on your gateway class)</p><p>You need to enable Inference Extension on your GatewayClass. For instance, I’m using Istio here</p><p>istioctl install --set values.pilot.env.ENABLE_GATEWAY_API_INFERENCE_EXTENSION=true -y</p><p>For other gateway class, please google ‘&lt;your gateway class name&gt; enable inference extension’</p><p>Note: If you’re running on restricted Kubernetes cluster like on Killercoda, kindly restrict your resources allocation else your Istio deployment will be stuck:</p><pre>istioctl install --set profile=minimal \<br>  --set values.pilot.env.SUPPORT_GATEWAY_API_INFERENCE_EXTENSION=true \<br>  --set values.pilot.env.ENABLE_GATEWAY_API_INFERENCE_EXTENSION=true \<br>  --set values.pilot.resources.requests.memory=512Mi \<br>  --set values.pilot.resources.requests.cpu=100m \<br>  -y</pre><p>2. Deploy the EPP and InferencePool with Helm (easiest)</p><pre>git clone https://github.com/llm-d/llm-d-router.git &amp;&amp; cd llm-d-router<br><br>helm dependency build ./config/charts/llm-d-router-gateway<br><br>helm -n inf install vllm-sim-demo ./config/charts/llm-d-router-gateway \<br>  --set router.modelServers.matchLabels.app=vllm-sim-demo \<br>  --set provider.name=istio \<br>  --set router.epp.resources.requests.cpu=500m \<br>  --set router.epp.resources.requests.memory=512Mi \<br>  --set router.epp.resources.limits.memory=1Gi</pre><p>Check: k -n inf get infpool</p><p>3. Modify HTTPRoute to reference the infpool to the backendRef</p><pre>apiVersion: gateway.networking.k8s.io/v1<br>kind: HTTPRoute<br>metadata:<br>  name: inf-route<br>  namespace: inf<br>spec:<br>  hostnames:<br>  - kyle-is-not-handsome.com<br>  parentRefs:<br>  - group: gateway.networking.k8s.io<br>    kind: Gateway<br>    name: inf-gtw<br>  rules:<br>  - backendRefs:<br>####### THIS PARTTTT<br>    - group: inference.networking.k8s.io<br>      kind: InferencePool<br>      name: vllm-sim-demo<br>      port: 8000<br>      weight: 1<br>########<br>    filters:<br>    - type: URLRewrite<br>      urlRewrite:<br>        path:<br>          replacePrefixMatch: /<br>          type: ReplacePrefixMatch<br>    matches:<br>    - path:<br>        type: PathPrefix<br>        value: /inference</pre><p>4. Test again the gateway</p><pre>curl -s kyle-is-not-handsome.com/inference/v1/models   | jq<br>{<br>  &quot;object&quot;: &quot;list&quot;,<br>  &quot;data&quot;: [<br>    {<br>      &quot;id&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>      &quot;object&quot;: &quot;model&quot;,<br>      &quot;created&quot;: 1790344406,<br>      &quot;owned_by&quot;: &quot;vllm&quot;,<br>      &quot;root&quot;: &quot;Qwen/Qwen3.5-397B-A17B&quot;,<br>      &quot;parent&quot;: null,<br>      &quot;max_model_len&quot;: 1024<br>    }<br>  ]<br>}</pre><p>It worked! You’ve setup an North-South Inference system with Gateway API and Inference Pool on Kubernetes ⭐️️️️</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=68b948f2d721" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[[CKNE Exam Prep — Hands On] 4.4 Implementing Pod-level Authentication and Authorization (Part 3)]]></title>
            <link>https://medium.com/@kylelzk/ckne-exam-prep-hands-on-4-4-implementing-pod-level-authentication-and-authorization-part-3-0e53a7e9b675?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/0e53a7e9b675</guid>
            <category><![CDATA[ckne]]></category>
            <category><![CDATA[authorization-policy]]></category>
            <category><![CDATA[kubernetes-networking]]></category>
            <category><![CDATA[istio]]></category>
            <category><![CDATA[kubernetes-rbac]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 16:09:30 GMT</pubDate>
            <atom:updated>2026-09-24T16:17:31.945Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Part 3 — Pod-level Authorization (AuthorizationPolicy)</blockquote><h4>Lab — Authorization with Istio</h4><p>Lab Link: <a href="https://killercoda.com/lorenzo-g/scenario/security-authorization-http-traffic">https://killercoda.com/lorenzo-g/scenario/security-authorization-http-traffic</a></p><p>Basically this lab is about setting up AuthorizationPolicy (abbreviation ap , a CRD from istio), to setup the authorization layer between pods.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jGcwpcXWDCaus91ydLfAIw.png" /><figcaption>Authorization</figcaption></figure><p>There’re 3 deployments on default ns: tester, booking, notification</p><p>(tbh I think tester pod is not necessary to show ap as you could just do curl -X POST &lt;clusterIP of the booking service&gt;/book , but it’s fine)</p><h4>FIRST OF ALL, are these pods managed by Istio?</h4><pre>root@controlplane:~$ k get ns --show-labels | grep inject <br>default              Active   4d23h   istio-injection=enabled,kubernetes.io/metadata.name=default</pre><p>so yes lorrr (Malaysian slang), istio-injection is enabled on the default namespace, meaning — —every pods in default ns will be automatically injected an istio sidecar on it, meaning — —managed by Istio.</p><h4>How to interact with these services? What HTTP Method and Path?</h4><blockquote>I got these from <em>k logs deploy &lt;deploy name&gt;</em></blockquote><pre>root@controlplane:~$ k logs deploy/notification-service-v1 <br>...<br><br>[GIN-debug] POST   /notify                   --&gt; notification-service/controller.Notify (3 handlers)<br><br>root@controlplane:~$ k logs deploy/booking-service-v1 <br>...<br>[GIN-debug] GET    /bookings                 --&gt; booking-service/controller.GetBookings (3 handlers)<br>[GIN-debug] POST   /book                     --&gt; booking-service/controller.PostBooking (3 handlers)</pre><p>notification svc:<br>POST /notify (invoked automatically from POST /book, or can be invoked directly)</p><p>booking service:</p><p>GET /bookings</p><p>POST /book (which in turn do a POST /notify to notification svc)</p><h4>Let’s play around, try these out:</h4><p>Collect inputs:</p><pre>k get svc ### get IPs &amp; NAME<br>curl &lt;booking-svc-IP&gt;/bookings<br>curl -X POST &lt;booking-svc-IP/book<br>curl &lt;booking-svc-IP&gt;/bookings # AGAIN<br><br>curl -X POST &lt;noti-svc-IP/notify<br><br>k exec tester -- curl &lt;booking-svc-name&gt;/bookings<br>k exec tester -- curl -X POST &lt;booking-svc-name&gt;/book<br>k exec tester -- curl -X POST &lt;noti-svc-name&gt;/notify</pre><blockquote>Alright, you’ve pretty much understood the situation, let’s go</blockquote><h4>1. First of all, setup a allow nothing AP</h4><pre>apiVersion: security.istio.io/v1<br>kind: AuthorizationPolicy<br>metadata:<br>  generation: 1<br>  name: allow-nothing<br>  namespace: default<br>spec: {}</pre><p>So now, it’d apply to default namespace, and everything is denied from RBAC, Authorization perpsective.</p><p>Note: don’t confuse with ‘deny-all’ Network Policy, that’s networking layer; These are Authorization layer enforced from Istio sidecar</p><p>Play around agn — everything fails</p><pre>root@controlplane:~$ curl &lt;booking svc IP&gt;/bookings<br>RBAC: access denied<br>root@controlplane:~$ curl -X POST &lt;notification-svc-IP&gt;/notify<br>RBAC: access denied<br>root@controlplane:~$ curl -X POST &lt;booking svc IP&gt;/book<br>RBAC: access denied</pre><h4>2. Setup AP to enable POST to booking deployment</h4><pre>apiVersion: security.istio.io/v1<br>kind: AuthorizationPolicy<br>metadata:<br>  namespace: default<br>  name: booking-service-policy<br>spec:<br>  selector:<br>    matchLabels:<br>      app: booking-service<br>  action: ALLOW<br>  rules:<br>  - to:<br>    - operation:<br>        methods: [&quot;POST&quot;]</pre><p>Validate</p><pre>$ curl -X POST 10.101.123.84/book<br>Booking could not be placed, notification service returned HTTP status=403</pre><p>Now you can POST to booking service, but notification part failed.</p><h4>3. Setup AP to enable POST to notification deployment, scoped to a specific booking Service Account (SA)</h4><p><strong>3.1 Create SA</strong></p><p>k create sa booking-service-account</p><p><strong>3.2 Setup SA on booking deployment</strong></p><p>k edit deploy &lt;booking deployment name&gt;</p><p>Then addserviceAccountName: booking-service-account to spec.template.spec</p><p><strong>3.3 Setup AP on notification deployment, scoped to SA</strong></p><pre>apiVersion: security.istio.io/v1<br>kind: AuthorizationPolicy<br>metadata:<br>  name: notification-service-policy<br>  namespace: default<br>spec:<br>  action: ALLOW<br>  rules:<br>  - from:<br>    - source:<br>        principals: [&quot;cluster.local/ns/default/sa/booking-service-account&quot;]<br>  selector:<br>    matchLabels:<br>      app: notification-service</pre><p><strong>3.4 Validate</strong></p><pre>$ curl -X POST 10.101.123.84/book<br>{&quot;id&quot;:&quot;a15c9656-6331-4902-b4f4-8ee85dbcecec&quot;,&quot;notification&quot;:[&quot;EMAIL&quot;]}</pre><h4>DONE the LAB, what’s next?</h4><p>If you still have time on the lab, try these:</p><ul><li>Setup AP to allows GET , scoped to /bookings path on booking deployments, so you can curl &lt;booking IP&gt;/bookings</li><li>By default, the booking deployment can’t do curl, i.e. you can’t k exec deploy/booking-name -- curl -X POST notification-service-name/notify , curl is not found there. So, how about use k debug &lt;booking pod&gt; with curlimages/curl , then curl to notification svc?</li></ul><p>That’s it! Feel free to leave in comments of your understanding or any experiments you’ve tried. I’ll reply to you for any questions :))</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0e53a7e9b675" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[[CKNE Exam Prep] 4.2 — Implementing Node and Pod Level Encryption]]></title>
            <link>https://medium.com/@kylelzk/ckne-exam-prep-4-2-implementing-node-and-pod-level-encryption-b68e6b7af7e4?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/b68e6b7af7e4</guid>
            <category><![CDATA[kubernetes-networking]]></category>
            <category><![CDATA[encryption]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[cilium]]></category>
            <category><![CDATA[ckne]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 15:48:26 GMT</pubDate>
            <atom:updated>2026-09-24T15:48:26.895Z</atom:updated>
            <content:encoded><![CDATA[<p>In k8s, you can encrypt traffic at two layers. Cilium can do it at L3/L4 with WireGuard or IPsec, which is transparent to apps and needs no code changes.</p><p>Istio can do it at L7 with mTLS, which adds workload identity (SPIFFE) and authorization.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L1x3zOj_hp13yj81hT7BIw.png" /><figcaption>Encryption</figcaption></figure><h3>Option 1: Wireguard (Pod + Node level)</h3><ul><li>Supports pod &lt;-&gt; pod, and node &lt;-&gt; node (host network) traffic</li><li>Prerequisites: kernel 5.6 or newer (or a WireGuard backport), and UDP 51871 open between all nodes.</li></ul><pre>  --set encryption.enabled=true \<br>  --set encryption.type=wireguard \<br>  --set encryption.nodeEncryption=true \<br>  --set encryption.strictMode.enabled=true \<br>  --set encryption.strictMode.cidr=&quot;10.0.0.0/8&quot; \<br>  --set encryption.strictMode.allowRemoteNodeIdentities=true</pre><ul><li><strong>encryption.enabled and </strong><strong>type=wireguard</strong> encrypt pod-to-pod traffic across nodes.</li><li><strong>nodeEncryption=true</strong> adds node-to-node and pod-to-node traffic. You</li><li><strong>strictMode</strong> drops any unencrypted pod traffic within that CIDR instead of silently sending it in cleartext. Set the CIDR to your pod CIDR.</li></ul><p>then k -n kube-system rollout restart ds/cilium to apply changes</p><p>Validate:</p><pre>kubectl -n kube-system exec ds/cilium -- cilium-dbg status | grep Encryption<br># Encryption: Wireguard [NodeEncryption: Enabled, cilium_wg0 (Pubkey: ..., Port: 51871, Peers: N)]<br><br>kubectl -n kube-system exec ds/cilium -- cilium-dbg encrypt status</pre><p>To confirm on the wire, run</p><p>tcpdump -ni &lt;primary host interface name, eg. enp1s0&gt; udp port 51871</p><p>on a node while two pods on different nodes talk.</p><pre>nginx → lxc → cilium_wg0 → [WireGuard encrypts] → enp1s0 → network<br>                                 ↑                   ↑<br>                      plaintext ends here    first time enp1s0<br>                                             touches the packet</pre><p>You should see only WireGuard UDP, and tcpdump -ni eth0 tcp port &lt;app-port&gt; should show nothing crossing between nodes.</p><p>Related concepts:<br>VXLAN, pod traffic, NIC interfaces</p><p>TODO: further improve this with practical exercise</p><p><a href="https://claude.ai/share/3773bc5e-7e4c-4eab-b293-656196584d84">https://claude.ai/share/3773bc5e-7e4c-4eab-b293-656196584d84</a></p><h3>Option 2: IPsec (pod-level only)</h3><p>Use when you need FIPS-validated crypto or can’t run WireGuard.</p><p>(Author note: I don’t really feel like covering this section, kindly search yourself 😆)</p><p>Basically you need to generate a secret named<em> cilium-ipsec-keys</em> in kube-system namespace first, before changing <em>encryption.type=ipsec </em>on Cilium.</p><h3>Option 3: Istio mTLS</h3><p>For this, I’d cover in &lt;<em>4.4 — Implementing Pod-level Authentication and Authorization&gt;</em></p><p>Yay, happy reading :)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b68e6b7af7e4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[[CKNE Exam Prep] 2.3 — Customizing CoreDNS for Services]]></title>
            <link>https://medium.com/@kylelzk/ckne-exam-prep-2-3-customizing-coredns-for-services-8886efd53d47?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/8886efd53d47</guid>
            <category><![CDATA[kubernetes-dns]]></category>
            <category><![CDATA[ckne]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[coredns]]></category>
            <category><![CDATA[kubernetes-networking]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 06:14:41 GMT</pubDate>
            <atom:updated>2026-09-26T16:25:51.311Z</atom:updated>
            <content:encoded><![CDATA[<p>When you install Kubernetes, there’s a CoreDNS deployment and service got deployed along. This is the reason why you could ping service name directly in the same namespace, or ping &lt;svc name&gt;.&lt;ns&gt; for cross NS to reach the correspondig services.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BRXDA0JMc65DLKy5TRpAog.png" /><figcaption>CoreDNS</figcaption></figure><h4>DNS Resolution and /etc/resolv.conf</h4><p>CoreDNS add a /etc/resolv.conf file directly inside every of your pod, it looks like this</p><pre>k run test-pod --image=nginx <br>pod/test-pod created<br><br>k exec test-pod -- cat /etc/resolv.conf<br>search default.svc.cluster.local svc.cluster.local cluster.local<br>nameserver 10.96.0.10<br>options ndots:5</pre><p>Here, 10.96.0.10 is the CoreDNS Service IP, and search means it’d append those one by one for your DNS search, eg. ping a-service-name , and the DNS would turn that into ping a-service-name.default.svc.cluster , which trying to find a service named like that in your default namespace; likewise, ping svc-a.ns-a would try to reach that domain in default namespace first, if not found, then add svc.cluster.local (the 2nd next search), reaching a service namesvc-a in namespace ns-a . Hope this part is understandable, if not, try to copy-and-paste above code to GenAI to fully understand this part. (yay everything readers don’t understand just push to GenAI)</p><h4>Ehh Kyle, how about ‘options ndots:5&#39;?</h4><p>First of all, this /etc/resolv.conf file can be modified from spec.dnsConfig attribute in pod. For instance:</p><pre>apiVersion: v1<br>kind: Pod<br>metadata:<br>  labels:<br>    run: netshoot<br>  name: netshoot-ndot-1<br>spec:<br>  containers:<br>  - image: nicolaka/netshoot<br>    name: netshoot<br>    resources: {}<br>    command: [&quot;sleep&quot;, &quot;infinity&quot;]<br>  dnsPolicy: ClusterFirst<br>  dnsConfig:<br>    options:<br>      - name: ndots<br>        value: &quot;1&quot;</pre><p>Above pod yaml will create pod, withoptions ndots:1 in /etc/resolv.conf</p><pre>k -n dns1 exec netshoot-ndot-1 -- cat /etc/resolv.conf<br><br>search dns1.svc.cluster.local svc.cluster.local cluster.local<br>nameserver 10.96.0.10<br>options ndots:1</pre><p>To demo what ndots does, I’ve these k8s resources:</p><pre>$ k -n dns1 get po<br>NAME              READY   STATUS    RESTARTS   AGE<br>netshoot-ndot-1   1/1     Running   0          88s<br>netshoot-ndot-2   1/1     Running   0          88s<br><br>$ k -n dns2 get svc<br>NAME    TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE<br>nginx   ClusterIP   10.200.0.239   &lt;none&gt;        80/TCP    23s</pre><p>so I’ve 2 pods in dns1 ns, and 1 service at dns2 ns</p><p>Start with conclusion: The pod with ndots:2is able resolve nginx.dns2 DNS (the nginx service at dns2 namespace), whereas pod with ndots:1 cannot, i.e.</p><pre>##### ON POD WITH ndots:1<br><br>$ k -n dns1 exec netshoot-ndot-1 -- nslookup nginx.dns2<br>Server:  10.96.0.10<br>Address: 10.96.0.10#53<br><br>** server can&#39;t find nginx.dns2: NXDOMAIN<br><br>command terminated with exit code 1<br><br>##### ON POD WITH ndots:2<br><br>$ k -n dns1 exec netshoot-ndot-2 -- nslookup nginx.dns2<br>;; Got recursion not available from 10.96.0.10<br>;; Got recursion not available from 10.96.0.10<br>Server:  10.96.0.10<br>Address: 10.96.0.10#53<br><br>Name: nginx.dns2.svc.cluster.local<br>Address: 10.200.0.239<br>;; Got recursion not available from 10.96.0.10<br><br>$ k -n dns1 exec netshoot-ndot-2 -- nslookup nginx.dns2.svc<br>Server:  10.96.0.10<br>Address: 10.96.0.10#53<br><br>** server can&#39;t find nginx.dns2.svc: NXDOMAIN<br><br>command terminated with exit code 1</pre><p>Above is the usage of ndots , so:</p><p>nginx.dns2 domain has 1 dot, lower than your pod’s ndotsattribute, hence it wouldn’t iterate over the ‘search’ domain of pod1, but it worked on pod2. Likewise, nginx.dns2.svc domain has 2 dots, not lower than your pod’s ndots , hence it wouldn’t iterate over search</p><p>Is it too confused? Copy my codeblock above and ask AI, then explain in the comments section 😆 I’ll definitely reply to your explanation 👍👍</p><p>Pretty much for configuration for the /etc/resolv.conf file in pod, you can overwrite it from spec.dnsConfig .</p><h4>Corefile — The configuration of CoreDNS</h4><p>Next, I want to demonstrate how to configure Corefile, the configuration behind CoreDNS.</p><p>So CoreDNS is a set of deployment + service, the deployment is backed by configmap.</p><pre>$ k -n kube-system get deploy coredns -oyaml<br>...<br>volumes:<br>      - configMap:<br>          defaultMode: 420<br>          items:<br>          - key: Corefile<br>            path: Corefile<br>          name: coredns<br>        name: config-volume</pre><p>You can modify the cm, then rollout the coredns deployment to get that into effect. For instance:</p><p>I’ve a service with deployment</p><pre>$ k get svc -n demo<br>NAME       TYPE        CLUSTER-IP    EXTERNAL-IP   PORT(S)   AGE<br>demo-web   ClusterIP   10.200.0.61   &lt;none&gt;        80/TCP    27h<br></pre><p>when I curl the service IP, it returns this</p><pre>$ curl 10.200.0.61<br>&lt;!doctype html&gt;<br>&lt;html&gt;<br>  &lt;head&gt;&lt;title&gt;Internal demo&lt;/title&gt;&lt;/head&gt;<br>  &lt;body&gt;<br>    &lt;h1&gt;Hello from demo-web!&lt;/h1&gt;<br>    &lt;p&gt;If you see this, google.com was rewritten by CoreDNS to demo-web.demo.svc.cluster.local&lt;/p&gt;<br>  &lt;/body&gt;<br>&lt;/html&gt;</pre><p>Basically a simple website.</p><p>So, now IN THE CLUSTER, I want google.com DNS resolution, to this service instead of the public google.com. I can add this inside the CoreDNS configmap</p><pre>$ k -n kube-system edit cm coredns<br><br>....<br>        hosts {<br>          10.200.0.61 google.com www.google.com<br>          ttl 30<br>          fallthrough<br>        }<br><br><br>        errors<br>        health {<br>           lameduck 5s<br>        }<br>        ready<br>###### ADD THIS SECTION<br>        hosts {<br>          10.200.0.61 google.com www.google.com<br>          ttl 30<br>          fallthrough<br>        }<br>###### <br>        kubernetes cluster.local in-addr.arpa ip6.arpa {<br>           pods insecure<br>           fallthrough in-addr.arpa ip6.arpa<br>           ttl 30<br>        }<br>....</pre><p>Then you restart the coredns deployment for this new config to take into effects.</p><pre>k -n kube-system rollout restart deploy/coredns</pre><p>Try google.com inside the cluster</p><pre>$ k -n dns1 exec netshoot-ndot-2 -- curl -s google.com<br>&lt;!doctype html&gt;<br>&lt;html&gt;<br>  &lt;head&gt;&lt;title&gt;Internal demo&lt;/title&gt;&lt;/head&gt;<br>  &lt;body&gt;<br>    &lt;h1&gt;Hello from demo-web!&lt;/h1&gt;<br>    &lt;p&gt;If you see this, google.com was rewritten by CoreDNS to demo-web.demo.svc.cluster.local&lt;/p&gt;<br>  &lt;/body&gt;<br>&lt;/html&gt;</pre><p>toda! google.com is now linking to a custom deployment in your cluster</p><p>That’s pretty much conclude my preparation for CoreDNS for CKNE exam. Feel free to leave your comments for anything you’d like to add :)))</p><p>References:<br><a href="https://arn-ob.medium.com/how-coredns-works-a84e90e6cb37">https://arn-ob.medium.com/how-coredns-works-a84e90e6cb37</a><br><a href="https://oneuptime.com/blog/post/2026-02-09-coredns-kubernetes-plugin-discovery/view">https://oneuptime.com/blog/post/2026-02-09-coredns-kubernetes-plugin-discovery/view</a></p><p><a href="https://coredns.io/2017/05/08/custom-dns-entries-for-kubernetes/">https://coredns.io/2017/05/08/custom-dns-entries-for-kubernetes/</a></p><p><a href="https://github.com/coredns/deployment/blob/master/kubernetes/Scaling_CoreDNS.md">https://github.com/coredns/deployment/blob/master/kubernetes/Scaling_CoreDNS.md</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8886efd53d47" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[[CKNE Exam Prep] 4.1 Securing Traffic with Network Policies]]></title>
            <link>https://medium.com/@kylelzk/ckne-exam-prep-4-1-securing-traffic-with-network-policies-3cad5fd1c42c?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/3cad5fd1c42c</guid>
            <category><![CDATA[kubernetes-network-policy]]></category>
            <category><![CDATA[ckne]]></category>
            <category><![CDATA[networking]]></category>
            <category><![CDATA[kubernetes]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Tue, 22 Sep 2026 14:30:52 GMT</pubDate>
            <atom:updated>2026-09-22T14:35:37.912Z</atom:updated>
            <content:encoded><![CDATA[<p>For CKNE exam, I assume you should able to know how to setup Network Policy to restrict traffic from or to pod / namespace / external IP.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TXcTFm-OwOLxEQ8BXWaLgw.png" /><figcaption>NP</figcaption></figure><p>Probably one of the common pitfalls could be forgetting to allow pods to Egress to CoreDNS pod. Basically when setting up NP with Egress type, you should setup another egress rule to allow access to kube-dns service IP, port 53 for TCP and UDP. (if forgot, you can check k run svc -A , to get the IP and ports for kube-dns)</p><p>Here are some of the practice labs for Network Policy, these should be the basics of NP before coming into CKNE prep.</p><ul><li><a href="https://killercoda.com/kylelaw/course/netpol/egress-and-the-dns-trap">https://killercoda.com/kylelaw/course/netpol/egress-and-the-dns-trap</a></li><li><a href="https://killercoda.com/saiyampathak/course/cnpe/04-networkpolicy-tenant-isolation">https://killercoda.com/saiyampathak/course/cnpe/04-networkpolicy-tenant-isolation</a></li><li><a href="https://killercoda.com/omkar-shelke25/scenario/NetworkPolicy-Cross-Namespace">https://killercoda.com/omkar-shelke25/scenario/NetworkPolicy-Cross-Namespace</a></li></ul><p>Also Network Policy can’t be created imperatively. Hence you’ve to be familiar in creating NP with references to Kubernetes official documentation / k explain netpol (netpol is abbreviation of networkpolicies, by checking with k api-resources | grep netpol )</p><p>Another note on NP is that, NP doesn’t get applied on Flannel CNI. Without CNI, NP is just a redundant Kubernetes object.</p><p>I guess I’d wrap up prep for this section. There are further underlying concepts like how it get applied with iptables, or Cilium L7 Network Policy to further restrict traffic based on HTTP action, but I wouldn’t dive into that deep unless I’ve finished covering other areas.</p><p>Learn enough to be dangerous and practical, use your time wisely :))</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3cad5fd1c42c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[[CKNE Exam Prep] 1.1 — Installing and Configuring CNI Plugins]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@kylelzk/ckne-exam-prep-1-1-installing-and-configuring-cni-plugins-e6e11abe8386?source=rss-e060f219712c------2"><img src="https://cdn-images-1.medium.com/max/2600/1*BOaDPCyHdbOdZV7FJrGP1w.png" width="3597"></a></p><p class="medium-feed-snippet">Kubernetes relies on CNI for the underlying Networking.</p><p class="medium-feed-link"><a href="https://medium.com/@kylelzk/ckne-exam-prep-1-1-installing-and-configuring-cni-plugins-e6e11abe8386?source=rss-e060f219712c------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@kylelzk/ckne-exam-prep-1-1-installing-and-configuring-cni-plugins-e6e11abe8386?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/e6e11abe8386</guid>
            <category><![CDATA[ckne]]></category>
            <category><![CDATA[kubernetes-networking]]></category>
            <category><![CDATA[cni]]></category>
            <category><![CDATA[kubernetes-cni]]></category>
            <category><![CDATA[kubernetes]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Tue, 22 Sep 2026 14:06:16 GMT</pubDate>
            <atom:updated>2026-09-22T14:08:49.738Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Exam Review — Nvidia Certified Professional — AI Operations with Labs (NCP-AIOL)]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@kylelzk/exam-review-nvidia-certified-professional-ai-operations-with-labs-ncp-aiol-22bcc042ca20?source=rss-e060f219712c------2"><img src="https://cdn-images-1.medium.com/max/917/1*IJetYQbGD_APYGhqEW2s9Q.png" width="917"></a></p><p class="medium-feed-snippet">Nvidia has recently changed their NCP-AIO exam to NCP-AIOL , changing format from pure 60+ multiple-choice questions to 35 MCQ + 3 Labs&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@kylelzk/exam-review-nvidia-certified-professional-ai-operations-with-labs-ncp-aiol-22bcc042ca20?source=rss-e060f219712c------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@kylelzk/exam-review-nvidia-certified-professional-ai-operations-with-labs-ncp-aiol-22bcc042ca20?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/22bcc042ca20</guid>
            <category><![CDATA[ncp-aiol]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-infrastructure]]></category>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[ai-operations]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Sun, 28 Jun 2026 01:05:02 GMT</pubDate>
            <atom:updated>2026-06-28T01:05:02.696Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Exam Review — NVIDIA-Certified Associate: AI Infrastructure and Operations (NCA-AIIO)]]></title>
            <link>https://medium.com/@kylelzk/exam-review-nvidia-certified-associate-ai-infrastructure-and-operations-nca-aiio-5920466a726a?source=rss-e060f219712c------2</link>
            <guid isPermaLink="false">https://medium.com/p/5920466a726a</guid>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[nvidia-gpu]]></category>
            <category><![CDATA[nca-aiio]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-infrastructure]]></category>
            <dc:creator><![CDATA[Kyle Law]]></dc:creator>
            <pubDate>Tue, 30 Sep 2025 05:03:00 GMT</pubDate>
            <atom:updated>2025-09-30T05:05:30.700Z</atom:updated>
            <content:encoded><![CDATA[<h3>[Exam Review] NVIDIA-Certified Associate: AI Infrastructure and Operations (NCA-AIIO)</h3><p>As NVIDIA becomes the largest company in the world (USD 4.33 trillion market cap as of 28 Sep 2025), it’s simply too big to ignore as an infrastructure professional.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tmNSXtTQvwaN6gpi0DNRsA.png" /><figcaption>NCA-AIIO cert</figcaption></figure><p>I decided to give its associate-level certification a shot and passed with a decent score.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oAOUgxfSrAV_rMj7iLgz6A.png" /><figcaption>My score</figcaption></figure><p>Hereby writing this article to share my exam review and tips that could help you aced the exam.</p><h3><strong>It’s fairly easy</strong></h3><p>To be honest, I found this exam relatively easy, I managed to finish all 50 questions in about 15 minutes (the time limit is 60 minutes). The questions and options are straightforward — no long case studies. It’s basically: you either know it, or you don’t. That’s why it’s important to cover every topic listed in the exam guide.</p><h3>How I prepare?</h3><ol><li>The official exam guide — <a href="https://www.nvidia.com/en-us/learn/certification/ai-infrastructure-operations-associate/">https://www.nvidia.com/en-us/learn/certification/ai-infrastructure-operations-associate/</a></li><li>The official &lt;AI Infrastructure and Operations Fundamentals&gt; course, available at their website and <a href="https://www.coursera.org/learn/ai-infrastructure-operations-fundamentals">Coursera</a></li><li>Suggested Readings mentioned in the exam guide</li><li>Shameless plug — my curated exam prep course at Udemy available <a href="https://www.udemy.com/course/draft/6818103/?couponCode=69C2225FED43860213D2">here</a> with concise videos and quizzes to maximize your chances of passing the exam.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TZrKb2Z_yyoGhnybrOuRdQ.png" /><figcaption>My NCA-AIIO Course on Udemy</figcaption></figure><p><em>Use coupon ‘BESTPRICE’ for the cheapest pricing (available for until 3 Oct), or feel free to contact me on </em><a href="https://www.linkedin.com/in/kyle-lawzhunkhing/"><em>Linkedin</em></a><em> for coupon code if it expired :)</em></p><h3>Is there coupon discount for the exam?</h3><p>At this time of writing, I found a 20% coupon for Nvidia cert at <a href="https://www.youtube.com/watch?v=06SuQdgmhtg">https://www.youtube.com/watch?v=06SuQdgmhtg</a></p><p>Use promo code “DANIELDAN20” for 20% off</p><h3>Is it useful?</h3><p>At the time of writing, I haven’t seen any job opportunities specifically requiring the NCA-AIIO certification — at least not in my country — Malaysia.</p><p>That said, it was still a worthwhile attempt. Going through the materials gave me a stronger understanding of AI infrastructure, which I believe will become increasingly valuable as NVIDIA’s ecosystem grows.</p><p><em>Thanks for reading my article. Feel free to connect with me on </em><a href="https://www.linkedin.com/in/kyle-lawzhunkhing/"><em>Linkedin</em></a><em> and share your exam journey :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5920466a726a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>