Srinivas Sarkar

@devops_nk and the fun part is they dont trust eachother, every component needs to prove thier identity to eachother via certs(mTLS) before establishing connection.

Kubernetes Control Plane Summary: API Server – Entry point to the cluster; validates and processes all requests Scheduler – Assigns Pods to suitable worker nodes Controller Manager – Ensures desired state matches the actual state etcd – Stores the entire cluster state as key-value data Cloud Controller Manager (CCM) – Integrates Kubernetes with cloud provider services

@devops_nk Kubelet bahut kaam ki cheez hai bhai 🫪 that's the guy responsible to create those static pods in the first place.

Kubernetes Data Plane Summary: Kubelet → Manages Pod lifecycle and health. Kube-proxy → Handles networking and service traffic. Container Runtime → Provides the container execution environment.

@suryanox7 @cloclodma maxUnavailable: 0 would make sure to keep all the 5 pods alive until 5 more gets created while maxSurge > 0 ain't it? While maxUnavailable:2 would make sure to keep 3 pods alive while other 2 new pods gets ready, as per the requirement?

@cloclodma Use RollingUpdate with maxUnavailable 0 and maxSurge > 0. This ensures new Pods are added before old ones are removed, keeping at least 3 Pods running during rollout.

Kubernetes questions Your team requires zero-downtime updates. How do you configure a deployment to ensure at least 3 Pods stay alive during a rollout of 5 total replicas?

Kubernetes gudhala modda.

@G5II11 Game theory #1

يضحكني كيف الرجال توهم مصدومين من هالشي الحين وأنا عارفته من يوم كان عمري 15

@devops_nk Static Pods let the control plane bring itself up from nothing. Static Pods exist so Kubernetes can run its most critical components without depending on Kubernetes itself.

This question was asked in my last DevOps interview. What is Static Pods in Kubernetes ? A Static Pod is created and managed directly by the kubelet on a node, not through the Kubernetes API server. It works by placing a pod manifest file on the node, usually in: /etc/kubernetes/manifests The kubelet continuously monitors this folder and automatically creates or restarts the pod if needed. Where are Static Pods used ? - Mostly for control plane components in self-managed clusters (like kubeadm), because these components must run even if the API server is not fully available. Examples: • kube-apiserver • kube-scheduler • kube-controller-manager • etcd Important: kubelet starts them directly on the node, ensuring the control plane can bootstrap itself. Where Static Pods are NOT used ? In managed Kubernetes services like: • Amazon EKS • Google GKE • Azure AKS The control plane is managed by the cloud provider, so these components are not visible as static pods. On worker nodes you typically only see components like: • kube-proxy • CoreDNS • CNI plugins Save it for your next interview and follow for more.

CRI talks to kubelet via gRPC. CRI then calls CNI for networking. CSI is called by the kubelet separately for storage. The chain is: kubelet → CRI (run container) → CNI (give it a network) and kubelet → CSI (give it storage). Don't mix up who calls who.

Kubernetes has two kinds of machines: Control Plane (the brain) and Nodes (the workers). Don't confuse them.

@devops_nk Wake up to reality

Finally !

@immanuel_vibe this is awesome, thanks Immanuel!

Prepping for a DevOps / SRE interview? Don't sleep on this free q&a hub 👉 ewry.net/devops-sre-int… Its like a compact roadmap of what you actually need before interview day hits you Everything's grouped by real tech areas: - Ansible, AWS, Docker, GitHub Actions, GitLab CI/CD, Kubernetes, Linux, networking, Terraform, PostgreSQL - and more That's the whole point You study by stack, not by random trivia that goes nowhere. Here's how to actually get the most out of it: 1. Run topic drills, pick one area - say Kubernetes. List the core internals. Then walk through real stuff like readiness vs liveness probes, or ingress vs service mesh. Why would you pick one over the other? Think it through 2. Turn questions into cheatsheets. Write your own notes. Real-world tradeoffs - canary vs blue/green, circuit breakers, retries, timeouts. In your own words, not copy-paste 3. Mix basics with system thinking. For networking, don't just memorize ports. Trace a request through a 3-tier app and ask yourself - where can this actually break? 4. Connect concepts to reliability goals. Map IAM, VPCs, autoscaling to SLOs and error budgets. The goal is explaining why decisions get made, not just what they are 5. Practice talking through outages. Pick a "service is down" scenario. Walk through metrics -> logs -> traces -> recovery. That's literally what SRE interviews are testing for No fluff, no filler #interview_questions #devops #sre #cloud #kubernetes #k8s #terraform #linux #interview #aws

@__karnati 1. Slim images 2. Multi stage builds 3. Minimize layers 4. Proper caching

Your Docker image size is 2GB Your app image is 2GB. Deployments are getting slow. CI takes forever to build. How do you reduce the size of the image and improve deployment speed?

@awsdevelopers AWS

Reply to this tweet with "AWS" and we’ll tell you which AWS Service you are

@devops_nk NAT gateway, route table.

If you launched an EC2 instance in a private subnet and it cannot access the internet to install packages. What could be missing ?

@devops_nk Sudo = 💰

Only for Linux Lovers: Me: will you be my valentine ? She: no way Me: sudo will you be my valentine ? She: yes .. yes.. yes ! Let's go !

@__karnati Will surely explore it this weekend.

@random_techy Network adress translation, nice concept to work with

@TowardsAws 2025?

Want to level in Cloud in 2025?☁️ I am looking for 100 cloud builders who want to learn and build together. If you are interested write "CLOUD" in the comment and I will send you invitation.

@devops_nk #L12" target="_blank" rel="nofollow noopener">github.com/srinivassarkar… Check here: Access ECR via OIDC in GitHub Actions: 1. Create an OIDC provider in AWS IAM. 2. Create an IAM role with OIDC trust policy. 3. Configure GitHub Actions workflow.

Hi Everyone , Anyone implemented GitHub Actions workflow to access ECR via OIDC instead of AWS creds? If yes, pls share your GitHub repo URL. Planning to implement, need for reference.

@devops_nk Thanks 😼

Day 3 of making your DevOps Journey Easy 👇 Get FREE servers to practice real Linux & DevOps troubleshooting. 🐧 - Most people learn DevOps. - Very few actually practice troubleshooting. - Try SadServers a platform that gives you free, real Linux servers with intentionally broken scenarios. What you do: - Get access to a real server - SSH into it - Investigate logs, services, and configs - Fix real-world issues (services, disk, permissions, networking, Docker, Kubernetes) - No tutorials - No step-by-step answers. - Just pure troubleshooting exactly like production. 🐧 SadServer: sadservers.com #Linux #DevOps

@fromcodetocloud Traces – where did it go

Observability in DevOps in one-liners✅ Metrics – what is happening Logs – why it is happening Alarms – when it becomes a problem Alerts – who needs to act Latency – how slow it is Error rate – how broken it is Availability – is it up or not If you miss any one of these, you’re flying blind in production. Tools change, These fundamentals don’t.