Ravindra BagaleCourses & study guides मराठी

Ravi Sir's Cards › DevOps

Chapter 18: Kubernetes Workloads: Cards

मराठी cards · Open the chapter

Revision cards

Why Kubernetes: Compose doesn't survive a dead server. A single server will fail one day, usually at a bad time.
Revision card 1 of 15: Why Kubernetes: Compose doesn't survive a dead server · Download PNG · Open card page

Lesson: 18.1 Why Kubernetes — Compose does not survive a dead host

What Kubernetes solves that Docker alone doesn't. Docker runs containers on one host but doesn't manage a fleet.
Revision card 2 of 15: What Kubernetes solves that Docker alone doesn't · Download PNG · Open card page

Lesson: 18.2 What this solves that Docker alone does not

A Pod and its phase. Pod status is the first thing you read when something is wrong.
Revision card 3 of 15: A Pod and its phase · Download PNG · Open card page

Lesson: 18.3 A Pod, and the phase you can quote

Bare Pod (lab only): dies and stays dead. Beginners create Pods directly and lose the app when it dies.
Revision card 4 of 15: Bare Pod (lab only): dies and stays dead · Download PNG · Open card page

Lesson: 18.4 Bare Pod — lab only, dies and stays dead

ReplicaSet: keep N Pods. Something must keep the right number of Pods running.
Revision card 5 of 15: ReplicaSet: keep N Pods · Download PNG · Open card page

Lesson: 18.5 ReplicaSet — keep N Pods

ReplicationController: the older name. Old guides still show ReplicationController.
Revision card 6 of 15: ReplicationController: the older name · Download PNG · Open card page

Lesson: 18.6 ReplicationController — the older name

Deployment: the YAML you actually open. Most daily Kubernetes work is editing a Deployment.
Revision card 7 of 15: Deployment: the YAML you actually open · Download PNG · Open card page

Lesson: 18.7 Deployment — the YAML you actually open

DaemonSet: one per node. Some tools must run on every machine, like log or monitoring agents.
Revision card 8 of 15: DaemonSet: one per node · Download PNG · Open card page

Lesson: 18.8 DaemonSet — one per node

StatefulSet: stable name and disk. Databases need each copy to find its own data again after a restart.
Revision card 9 of 15: StatefulSet: stable name and disk · Download PNG · Open card page

Lesson: 18.9 StatefulSet — stable name and disk

Job and CronJob. Some work isn't a server; it runs, finishes and stops.
Revision card 10 of 15: Job and CronJob · Download PNG · Open card page

Lesson: 18.10 Job and CronJob

kubectl get, describe, logs, delete. Four commands answer most questions about a workload.
Revision card 11 of 15: kubectl get, describe, logs, delete · Download PNG · Open card page

Lesson: 18.11 kubectl get, describe, logs, delete

The Service spreads traffic across ready Pods. Users need one address even as Pods come and go.
Revision card 12 of 15: The Service spreads traffic across ready Pods · Download PNG · Open card page

Lesson: 18.12 The Service spreads traffic across ready Pods

Autoscaling: HPA. Traffic changes during the day. Fixed replicas waste money or fall over.
Revision card 13 of 15: Autoscaling: HPA · Download PNG · Open card page

Lesson: 18.13 Autoscaling — HPA, and two names to recognise

Workload safety basics. Defaults are not safe enough for production.
Revision card 14 of 15: Workload safety basics · Download PNG · Open card page

Lesson: 18.14 Safety basics

Lab: Deployment, not a bare Pod. Seeing self-healing once makes it stick.
Revision card 15 of 15: Lab: Deployment, not a bare Pod · Download PNG · Open card page

Lesson: 18.15 Lab — Deployment, not a bare Pod

Interview cards

What happens to your app if the server dies? With Compose on one VM: down. With Kubernetes and 2+ replicas: Pods are rescheduled on other nodes.
Interview card 1 of 15: What happens to your app if the server dies? · Download PNG · Open card page

Lesson: 18.1 Why Kubernetes — Compose does not survive a dead host

What is container orchestration? Placing, watching, scaling and updating containers across many machines automatically.
Interview card 2 of 15: What is container orchestration? · Download PNG · Open card page

Lesson: 18.2 What this solves that Docker alone does not

What does a Pod in Pending mean? It's accepted but not running yet: no node with space, or image still pulling, or volume not bound.
Interview card 3 of 15: What does a Pod in Pending mean? · Download PNG · Open card page

Lesson: 18.3 A Pod, and the phase you can quote

Why not run apps as bare Pods? No controller recreates them. If the Pod or node dies, the app is gone.
Interview card 4 of 15: Why not run apps as bare Pods? · Download PNG · Open card page

Lesson: 18.4 Bare Pod — lab only, dies and stays dead

What does a ReplicaSet do? Keeps a fixed number of matching Pods running, creating or deleting Pods as needed.
Interview card 5 of 15: What does a ReplicaSet do? · Download PNG · Open card page

Lesson: 18.5 ReplicaSet — keep N Pods

ReplicationController vs ReplicaSet? Both keep N Pods. ReplicaSet is newer, supports set-based selectors, and is used by Deployments.
Interview card 6 of 15: ReplicationController vs ReplicaSet? · Download PNG · Open card page

Lesson: 18.6 ReplicationController — the older name

What are the key fields of a Deployment? replicas, selector, template (Pod spec with containers), and strategy.
Interview card 7 of 15: What are the key fields of a Deployment? · Download PNG · Open card page

Lesson: 18.7 Deployment — the YAML you actually open

When do you use a DaemonSet? When one Pod per node is needed, like log collectors, monitoring agents or network plugins.
Interview card 8 of 15: When do you use a DaemonSet? · Download PNG · Open card page

Lesson: 18.8 DaemonSet — one per node

Why do databases use StatefulSets? Each Pod gets a stable name (db-0) and its own disk, so it finds its data after restart.
Interview card 9 of 15: Why do databases use StatefulSets? · Download PNG · Open card page

Lesson: 18.9 StatefulSet — stable name and disk

How do you run a nightly task in Kubernetes? A CronJob with a schedule creates a Job that runs a Pod to completion.
Interview card 10 of 15: How do you run a nightly task in Kubernetes? · Download PNG · Open card page

Lesson: 18.10 Job and CronJob

How do you debug a workload with kubectl? get to see status, describe for events, logs for app output, then fix the manifest.
Interview card 11 of 15: How do you debug a workload with kubectl? · Download PNG · Open card page

Lesson: 18.11 kubectl get, describe, logs, delete

Which Pods does a Service send traffic to? Pods matching its selector that are Ready. Not-ready Pods get no traffic.
Interview card 12 of 15: Which Pods does a Service send traffic to? · Download PNG · Open card page

Lesson: 18.12 The Service spreads traffic across ready Pods

How does the HPA work? It reads metrics (like CPU vs requests) and changes replicas between min and max.
Interview card 13 of 15: How does the HPA work? · Download PNG · Open card page

Lesson: 18.13 Autoscaling — HPA, and two names to recognise

What makes a Deployment production-safe? ≥2 replicas, probes, requests and limits, non-root, version tag, and a PodDisruptionBudget.
Interview card 14 of 15: What makes a Deployment production-safe? · Download PNG · Open card page

Lesson: 18.14 Safety basics

Show that Kubernetes self-heals. Create a Deployment with 3 replicas, delete a Pod, and watch a new one start.
Interview card 15 of 15: Show that Kubernetes self-heals. · Download PNG · Open card page

Lesson: 18.15 Lab — Deployment, not a bare Pod

Comparison cards

Kubernetes Deployment vs StatefulSet. Kubernetes runs apps as Pods.Deployment: runs N identical Pods. Any Pod can replace any other.StatefulSet: runs Pods that each keep a fixed name (db-0, db-1) and their own disk.
Comparison card 1 of 3: Kubernetes Deployment vs StatefulSet · Download PNG · Open card page

Lesson: 18.9 StatefulSet — stable name and disk

Deployment vs DaemonSet. Deployment: N Pods placed anywhere.DaemonSet: exactly one Pod on each node.
Comparison card 2 of 3: Deployment vs DaemonSet · Download PNG · Open card page

Lesson: 18.8 DaemonSet — one per node

Job vs CronJob. Job: run Pods once until success.CronJob: create a Job on a schedule.
Comparison card 3 of 3: Job vs CronJob · Download PNG · Open card page

Lesson: 18.10 Job and CronJob