🖼️ See the drawing + lab 🏠 Course home 🌿 Branch on GitHub ✏️ View source

🖼️ आकृती आणि labThe drawing + lab पूर्ण पानावर उघडा ↗Open full page ↗

🧑‍🏫 धडा 03 — Deployments: कडक वर्ग-प्रमुख

📍 तुम्ही इथे आहात: 13 पैकी धडा 03 · मागचा: lesson-02-pods · पुढचा: lesson-04-services

📦 या ब्रँचमध्ये काय आहे

धडे 01–02, आणि: Deployment — तुमचे pods योग्य संख्येने, योग्य version चालवत जिवंत ठेवणारा object. खऱ्या files:

k8s/deployment.yaml — Node API चा Deployment (2 replicas)
k8s/analytics-deployment.yaml — तीच पद्धत, Python service

🧒 5 वर्षांच्या मुलाला समजावल्यासारखे

शिक्षिका वर्ग-प्रमुखाला सांगतात: "झाडांना पाणी घालायला नेहमी 2 मुले असलीच पाहिजेत. नेहमी. कसे ते मला माहीत नाही." 🧑‍🏫🌱

वर्ग-प्रमुख हे गांभीर्याने घेतो:

एखादे मूल आजारी पडून घरी गेले? → वर्ग-प्रमुख लगेच नवे मूल पाठवतो. प्रश्न नाहीत.
कोणी चोरून तिसरे मूल आणले? → वर्ग-प्रमुख एकाला परत पाठवतो. बरोबर 2. नेहमी.
शिक्षिकांनी नियम बदलून "3 मुले" केला? → वर्ग-प्रमुख एक वाढवतो.

हाच Deployment. तुम्ही Kubernetes ला काय करायचे ते सांगत नाही — तुम्ही त्याला जग कसे दिसायला हवे ते सांगता ("school-api च्या 2 प्रती, version v5"), आणि वर्ग-प्रमुख वास्तव तुमच्या इच्छेशी जुळते ठेवण्यासाठी कायम काम करतो. Kubernetes मधली ही एकमेव सर्वात मोठी कल्पना: desired state (इच्छित स्थिती). तुम्ही जाहीर करता; तो जुळवून घेतो.

🗺️ आकृती

flowchart TB
    Y["🧑‍🏫 You declare:<br/>replicas: 2, image: v5"]
    D["Deployment<br/>school-api"]
    RS["ReplicaSet - the monitor<br/>keeps count = 2"]
    P1["🪑 Pod school-api-abc12"]
    P2["🪑 Pod school-api-def34"]
    P3["🪑 Pod school-api-xyz99<br/>auto-created replacement"]
    Y --> D --> RS
    RS --> P1
    RS --> P2
    P1 -. "💥 crashes" .-> P3
    RS -- "notices 1 missing,<br/>creates a new one" --> P3

❓ काय

Deployment जाहीर करतो: कोणती image, किती replicas, त्यांना update कसे करायचे. आतून तो एक ReplicaSet (खरा मोजणारा/वर्ग-प्रमुख) तयार करतो, आणि तो Pods तयार करतो.
Labels आणि selectors हे जोडणारे गोंद आहेत: ReplicaSet दावा करतो "app: school-api label असलेले pods माझे" — हाच धडा 02 मधला sticker.
Deployment → ReplicaSet → Pods. तुम्ही Deployment edit करता; बाकी आपोआप मागे येते.

🧠 Desired state → reconciliation → actual state

Desired state        replicas: 2          (what you wrote in git)
        ↓
Kubernetes continuously compares          (the controller loop, every few seconds, forever)
        ↓
Actual state         1 Pod                (one just crashed)
        ↓
creates another Pod
        ↓
Actual state = Desired state   2 Pods ✅

Kubernetes ही एक reconciliation (जुळवणी) system आहे. तुम्ही त्याला कधीच काय करायचे ते सांगत नाही — तुम्ही गोष्टी कशा असाव्यात ते सांगता, आणि controllers दरी बुजवतात, कायम. हा एकच loop Deployments, ReplicaSets, HPA (L09), Jobs (L18), StatefulSets (L22), operators (L25) आणि ArgoCD (L14) समजावतो — तोच loop, फक्त नवी टोपी.

🤔 का

धडा 02 मेलेला pod मेलेलाच राहतो यावर संपला. खऱ्या apps ना रात्री 3 वाजता तुमच्याशिवाय स्वतःहून बरे (self-heal) व्हावे लागते. Deployment मुळेच Kubernetes वाले लोक शांत झोपतात: crash झालेला pod, reboot झालेला node, मारलेला container — वर्ग-प्रमुख या सगळ्यांना बदलतो, कारण तुमच्या YAML ने सांगितले होते "2 असतील". तो एका वेळी एक pod बदलून zero-downtime upgrades (धडा 11) सुद्धा शक्य करतो.

🔧 कसे (या repo मध्ये)

k8s/deployment.yaml उघडा:

spec:
  replicas: 2                 # ← the promise: ALWAYS 2 desks
  selector:
    matchLabels:
      app: school-api         # ← "my pods wear this sticker"
  template:                   # ← lesson 02's pod blueprint
    metadata:
      labels:
        app: school-api       # ← the sticker itself (must match!)

image: IMAGE_PLACEHOLDER कडे लक्ष द्या — deploy च्या वेळी CircleCI त्यात खरा ECR image tag टाकतो (.circleci/config.yml पाहा).

🧪 करून पाहा

kubectl apply -f k8s/namespace.yaml          # the room first (lesson 05 explains)
kubectl apply -f k8s/configmap.yaml          # config (lesson 06 explains)

# For local play, use a public image instead of IMAGE_PLACEHOLDER:
sed 's|IMAGE_PLACEHOLDER|nginxdemos/hello:plain-text|; s|containerPort: 3000|containerPort: 80|' \
  k8s/deployment.yaml | kubectl apply -f -

kubectl -n school get pods                   # 2 pods, names like school-api-xxxxx

# 😈 Now be the chaos: delete one pod...
kubectl -n school delete pod -l app=school-api --field-selector=status.phase=Running --wait=false
kubectl -n school get pods -w                # ...watch the monitor replace it in seconds!  Ctrl+C to stop

🔧 या धड्यासाठी kubectl

kubectl get deploy · kubectl get rs · kubectl scale · kubectl rollout status

🔗 यावर आधारित: 🪪 AWS L12 (ASG म्हणजे 'computers साठीचा Deployment')

✅ तपासा — तुम्हाला काय दिसायला हवे

दोनपैकी एकावर kubectl delete pod → काही सेकंदांत kubectl get pods मध्ये NEW pod चे नाव आणि पुन्हा 2/2 दिसते — reconciliation प्रत्यक्ष काम करताना.

🧹 साफसफाई

local cluster: kubectl delete -f k8s/ (किंवा तुम्ही apply केलेली file) — EKS वर चालू राहिलेल्या प्रत्येक गोष्टीचे तासाप्रमाणे बिल लागते

⚠️ नेहमीच्या चुका

ReplicaSet वर replicas edit करणे — Deployment edit करा, RS फक्त त्याचा मोजणारा आहे
production मध्ये हाताने kubectl scale करणे आणि GitOps ते परत बदलेल हे विसरणे (L14)

⏭️ पुढे

आता pods स्वतःहून बरे होतात — पण त्यांची नावे आणि IPs प्रत्येक वेळी बदलतात. मग कोणी त्यांना खात्रीने शोधायचे कसे? Service — शाळेचे स्वागत-कक्ष (reception).

git checkout lesson-04-services

🧑‍🏫 Lesson 03 — Deployments: the strict class monitor

📍 You are here: Lesson 03 of 13 · Previous: lesson-02-pods · Next: lesson-04-services

📦 What's in this branch

Lessons 01–02, plus: the Deployment — the object that keeps your pods alive, in the right number, running the right version. Real files:

k8s/deployment.yaml — the Node API Deployment (2 replicas)
k8s/analytics-deployment.yaml — same pattern, Python service

🧒 Explain like I'm 5

The teacher tells the class monitor: "There must ALWAYS be 2 kids watering the plants. Always. I don't care how." 🧑‍🏫🌱

The monitor takes this seriously:

A kid goes home sick? → monitor instantly sends a new kid. No questions.
Someone sneaks in a 3rd kid? → monitor sends one back. Exactly 2. Always.
Teacher changes the rule to "3 kids"? → monitor adds one.

That's a Deployment. You don't tell Kubernetes what to do — you tell it what the world should look like ("2 copies of school-api, version v5"), and the monitor works forever to keep reality matching your wish. This is the single biggest idea in Kubernetes: desired state. You declare; it reconciles.

🗺️ Diagram

flowchart TB
    Y["🧑‍🏫 You declare:<br/>replicas: 2, image: v5"]
    D["Deployment<br/>school-api"]
    RS["ReplicaSet - the monitor<br/>keeps count = 2"]
    P1["🪑 Pod school-api-abc12"]
    P2["🪑 Pod school-api-def34"]
    P3["🪑 Pod school-api-xyz99<br/>auto-created replacement"]
    Y --> D --> RS
    RS --> P1
    RS --> P2
    P1 -. "💥 crashes" .-> P3
    RS -- "notices 1 missing,<br/>creates a new one" --> P3

❓ What

A Deployment declares: which image, how many replicas, how to update them. Under the hood it creates a ReplicaSet (the actual counter/monitor), which creates the Pods.
Labels & selectors are the glue: the ReplicaSet claims "my pods are the ones labeled app: school-api" — that's the sticker from lesson 02.
Deployment → ReplicaSet → Pods. You edit the Deployment; the rest follows.

🧠 Desired state → reconciliation → actual state

Desired state        replicas: 2          (what you wrote in git)
        ↓
Kubernetes continuously compares          (the controller loop, every few seconds, forever)
        ↓
Actual state         1 Pod                (one just crashed)
        ↓
creates another Pod
        ↓
Actual state = Desired state   2 Pods ✅

Kubernetes is a reconciliation system. You never tell it what to do — you tell it how things should be, and controllers close the gap, forever. This one loop explains Deployments, ReplicaSets, HPA (L09), Jobs (L18), StatefulSets (L22), operators (L25) and ArgoCD (L14) — same loop, new hat.

🤔 Why

Lesson 02 ended with a dead pod staying dead. Real apps must self-heal at 3 AM without you. The Deployment is why Kubernetes people sleep well: a crashed pod, a rebooted node, a killed container — the monitor replaces them all, because your YAML said "there shall be 2". It also unlocks zero-downtime upgrades (lesson 11) by swapping pods one at a time.

🔧 How (in this repo)

Open k8s/deployment.yaml:

spec:
  replicas: 2                 # ← the promise: ALWAYS 2 desks
  selector:
    matchLabels:
      app: school-api         # ← "my pods wear this sticker"
  template:                   # ← lesson 02's pod blueprint
    metadata:
      labels:
        app: school-api       # ← the sticker itself (must match!)

Note image: IMAGE_PLACEHOLDER — CircleCI stamps in the real ECR image tag at deploy time (see .circleci/config.yml).

🧪 Try it

kubectl apply -f k8s/namespace.yaml          # the room first (lesson 05 explains)
kubectl apply -f k8s/configmap.yaml          # config (lesson 06 explains)

# For local play, use a public image instead of IMAGE_PLACEHOLDER:
sed 's|IMAGE_PLACEHOLDER|nginxdemos/hello:plain-text|; s|containerPort: 3000|containerPort: 80|' \
  k8s/deployment.yaml | kubectl apply -f -

kubectl -n school get pods                   # 2 pods, names like school-api-xxxxx

# 😈 Now be the chaos: delete one pod...
kubectl -n school delete pod -l app=school-api --field-selector=status.phase=Running --wait=false
kubectl -n school get pods -w                # ...watch the monitor replace it in seconds!  Ctrl+C to stop

🔧 kubectl for this lesson

kubectl get deploy · kubectl get rs · kubectl scale · kubectl rollout status

🔗 Builds on: 🪪 AWS L12 (an ASG is 'a Deployment for computers')

✅ Verify — what you should see

kubectl delete pod one of the two → within seconds kubectl get pods shows a NEW pod name and 2/2 again — reconciliation in action.

🧹 Clean up

local cluster: kubectl delete -f k8s/ (or the file you applied) — on EKS, anything left running bills by the hour

⚠️ Common mistakes

editing replicas on the ReplicaSet — edit the Deployment, the RS is its counter
kubectl scale by hand in production and forgetting GitOps will revert it (L14)

⏭️ Next

Pods now heal themselves — but their names and IPs change every time. How does anyone find them reliably? The Service — the school reception desk.

git checkout lesson-04-services

← Previouspods Next →services

This page is the lesson's README from the lesson-03-deployments branch, shown here so the whole School stays on one site. Code files open on GitHub at the same branch.