Skip to content
Go back

Week 4: Orchestration Fundamentals

Edit page

🐳 Docker 360 β€” Week 4 of 16 | Orchestration: Swarm cluster, replicas, rolling updates aur self-healing

🎯 Learning Objective

Ab tak hum ne single machine par Docker chalaya. Lekin production me sawal hota hai: agar machine hi down ho jaye to? Traffic barh jaye to? Iska jawab hai orchestration β€” containers ko cluster me chalana, scale karna aur khud heal karna. Is week me tum Docker Swarm ka cluster banao ge, services ko replicate karo ge, zero-downtime rolling updates karo ge, aur dekho ge ke Swarm failed containers ko khud kaise wapas lata hai.

🎨 Visual Diagrams

Swarm cluster architecture

  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
  β”‚                    SWARM CLUSTER                        β”‚
  β”‚                                                         β”‚
  β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”      β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”                   β”‚
  β”‚  β”‚ MANAGER NODE β”‚      β”‚ WORKER NODE  β”‚                   β”‚
  β”‚  β”‚              β”‚      β”‚              β”‚                   β”‚
  β”‚  β”‚ β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”   β”‚      β”‚ β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”   β”‚                   β”‚
  β”‚  β”‚ β”‚ task 1 β”‚   β”‚      β”‚ β”‚ task 3 β”‚   β”‚                   β”‚
  β”‚  β”‚ β”‚ (web)  β”‚   β”‚      β”‚ β”‚ (web)  β”‚   β”‚                   β”‚
  β”‚  β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β”‚      β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β”‚                   β”‚
  β”‚  β”‚ β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”   β”‚      β”‚ β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”   β”‚                   β”‚
  β”‚  β”‚ β”‚ task 2 β”‚   β”‚      β”‚ β”‚ task 4 β”‚   β”‚                   β”‚
  β”‚  β”‚ β”‚ (web)  β”‚   β”‚      β”‚ β”‚ (web)  β”‚   β”‚                   β”‚
  β”‚  β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β”‚      β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β”‚                   β”‚
  β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜      β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜                   β”‚
  β”‚         β”‚                      β”‚                          β”‚
  β”‚         └────── Raft consensus β”€β”˜                          β”‚
  β”‚         (managers cluster state agree karte hain)         β”‚
  β”‚                                                         β”‚
  β”‚  SERVICE "web" (replicas: 4)                             β”‚
  β”‚   = desired state. Swarm isko har haal me                 β”‚
  β”‚     maintain karta hai β€” node down ho ya                  β”‚
  β”‚     container crash, naye tasks schedule                  β”‚
  β”‚     ho jate hain.                                         β”‚
  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

  Tum SERVICE (kya chahiye) declare karte ho,
  Swarm TASKS (asal containers) khud manage karta hai.

Rolling update β€” zero downtime kaise?

  Time ──────────────────────────────────────────────►

  Start:   [v1] [v1] [v1] [v1] [v1] [v1]   (6 replicas)

  Step 1:  [v1] [v1] [v1] [v1] [v1] [v2]   (1 update ho raha)
           update_config: parallelism 1

  Step 2:  [v1] [v1] [v1] [v1] [v2] [v2]   (purana stop, naya
                                            healthy hone par aage)

  Step 3:  [v1] [v1] [v1] [v2] [v2] [v2]

  ...
  Done:    [v2] [v2] [v2] [v2] [v2] [v2]   (koi downtime nahi!)

  Agar naya version healthcheck fail kare to update RUK jata hai β€”
  aur `service update --rollback` se purana version wapas.

Restart policy β€” self-healing flow

  Container crash ho gaya (exit code != 0)
              β”‚
              β–Ό
  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
  β”‚ Swarm task state:     β”‚
  β”‚  SHUTDOWN (failed)    β”‚
  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
              β”‚  restart policy check:
              β”‚  condition: on-failure
              β–Ό
  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
  β”‚ delay 5s, phir naya   │──► max-attempts cross?
  β”‚ task schedule karo    β”‚         β”‚ yes
              β”‚           β”‚         β–Ό
              β–Ό           β”‚   task: REJECTED
  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”    (manual fix chahiye)
  β”‚ naya container UP,    β”‚
  β”‚ desired replicas      β”‚
  β”‚ wapas poore!          β”‚
  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

🧠 Theory Deep Dive

Orchestration kyun zaroori hai?

Single docker run me ye problems hain:

Orchestrator desired state leta hai (β€œmujhe web ke 6 replicas chahiye, image v2 par”) aur cluster ko us state par rakhta hai β€” yehi β€œreconciliation loop” hai.

Docker Swarm vs Kubernetes β€” imandari se comparison

Docker SwarmKubernetes
Seekhne meBohat easy β€” Docker CLI hi kaafiLearning curve tez hai
Setupdocker swarm init β€” 1 commandkubeadm/k3s/managed service chahiye
ConceptsService, task, stackPod, Deployment, Service, Ingress…
Scaling--replicas NHPA, VPA, cluster autoscaler
EcosystemChhotaBohat bara (Helm, operators, GitOps)
Industry useChhote/medium setups, edgeLarge-scale production standard
StatusDocker me built-in, maintainedCNCF flagship project

Imandari wali baat: Swarm seekhna orchestration ke concepts (desired state, scheduling, rolling updates, service discovery) samajhne ka sab se tez tareeqa hai β€” aur ye concepts Kubernetes me bhi wahi hain, sirf naam aur tooling alag. Swarm production me bhi chalta hai jahan simplicity chahiye, lekin bari industry Kubernetes par hai. Dono seekhna faida deta hai.

Swarm ke core concepts

Rolling updates aur rollback

docker service update \
  --image nginx:1.27-alpine \
  --update-parallelism 2 \
  --update-delay 10s \
  --update-failure-action rollback \
  web

Iska matlab ye hai ke deployments me downtime ka dar khatam β€” aur agar kuch garbar ho to rollback ek command par.

πŸ’» Hands-On Lab

Lab: single machine kaafi hai β€” swarm init se ye khud manager + worker ban jata hai.

Lab A β€” Swarm cluster init

Step 1: Swarm initialize karo.

# Apni machine ka LAN IP nikalo
MYIP=$(ip -4 addr show eth0 2>/dev/null | grep -oP '(?<=inet\s)\d+(\.\d+){3}' | head -1)
echo "Using IP: $MYIP"
docker swarm init --advertise-addr "$MYIP"

βœ… Expected result: Swarm initialized: current node ... is now a manager. Sath me worker join karne ka token bhi milta hai.

Step 2: Cluster info dekho.

docker node ls
docker swarm join-token worker   # doosri machine hoti to ye command wahan chalti

βœ… Expected result: node ls me ek node Leader status ke sath. (Multi-node lab me token wali command worker machines par chalate.)

Lab B β€” Replicated service

Step 3: 3 replicas wali web service banao.

docker service create --name web --replicas 3 -p 8080:80 nginx:alpine
sleep 5
docker service ls
docker service ps web

βœ… Expected result: service ls me 3/3 replicas, service ps me 3 tasks Running state me. Teen containers β€” tum ne ek command chalayi, Swarm ne placement khud ki.

Step 4: Scale up/down karo.

docker service scale web=5
sleep 5
docker service ps web --format '{{.Name}} {{.CurrentState}}' | sort | uniq -c
curl -s -o /dev/null -w "%{http_code}\n" http://localhost:8080

βœ… Expected result: 5 tasks running, aur curl par 200. Routing mesh ki wajah se port 8080 har node par kaam karta hai.

Lab C β€” Rolling update aur rollback

Step 5: Rolling update chalao.

docker service update \
  --image nginx:1.27-alpine \
  --update-parallelism 1 \
  --update-delay 5s \
  web
docker service ps web

βœ… Expected result: Tasks ek ek karke Shutdown β†’ naya task Running hota hai. Beech me curl chalate raho:

for i in 1 2 3 4 5 6; do curl -s -o /dev/null -w "%{http_code} " http://localhost:8080; sleep 3; done; echo

βœ… Expected result: Sab 200 β€” update ke doran koi failed request nahi. Yehi zero-downtime deployment hai.

Step 6: Rollback karo.

docker service update --rollback web
sleep 8
docker service ps web --format '{{.Image}}' | sort | uniq -c

βœ… Expected result: Images wapas purani (nginx:alpine) par β€” rollback ek command me.

Lab D β€” Healthcheck + restart policy = self-healing

Step 7: Healthcheck wali service banao.

docker service create --name api \
  --replicas 2 \
  --health-cmd "wget --no-verbose --tries=1 --spider http://localhost/ || exit 1" \
  --health-interval 10s \
  --health-timeout 3s \
  --health-retries 2 \
  --restart-condition on-failure \
  --restart-delay 5s \
  --restart-max-attempts 3 \
  nginx:alpine
sleep 12
docker service ps api

βœ… Expected result: Dono tasks Running, aur inspect me health status healthy.

Step 8: Ek container ko zabardasti maar do β€” Swarm kya karta hai?

TASK_CONTAINER=$(docker ps --filter "name=api" --format '{{.ID}}' | head -1)
echo "Killing: $TASK_CONTAINER"
docker kill "$TASK_CONTAINER"
sleep 10
docker service ps api

βœ… Expected result: Mari hui task Failed/Shutdown me, aur uski jagah ek nayi task Running β€” Swarm ne desired state (2 replicas) khud restore kar di. Tumhe kuch nahi karna para. Yehi self-healing hai.

Step 9: Poori service ka health dekho.

docker service inspect api --format='{{json .Spec.TaskTemplate.ContainerSpec.Healthcheck}}' | head -c 200; echo
docker service ls

βœ… Expected result: Healthcheck config aur 2/2 replicas β€” cluster apni desired state par wapas.

Cleanup:

docker service rm web api
docker swarm leave --force
docker network prune -f

βœ… Expected result: Services removed, node swarm se nikal gaya. Machine wapas normal Docker mode me.

βœ… Checkpoint

Q1: Swarm me service aur task me kya farq hai? A: Service desired state ka elaan hai (β€œnginx ke 3 replicas chahiye”). Task us service ka ek chalta hua instance hai β€” asal me ek container. Ek service ke multiple tasks hote hain.

Q2: Rolling update me --update-parallelism 1 aur --update-delay ka kya role hai? A: Parallelism control karta hai ke ek waqt me kitne tasks update hon, delay har batch ke darmiyan waqfa deta hai taake naya task healthy ho jaye. Dono mil kar zero-downtime updates guarantee karte hain.

Q3: Container crash hone par Swarm khud kya karta hai? A: Restart policy (on-failure) ke mutabiq delay ke baad naya task schedule karta hai taake replicas ki desired count poori rahe. Tumhe manual docker run nahi karna parta β€” yehi reconciliation loop hai.

Q4: Kubernetes seekhne se pehle Swarm kyun useful hai? A: Orchestration ke core concepts β€” desired state, scheduling, service discovery, rolling updates, healthchecks β€” dono me same hain. Swarm inko Docker CLI se 10 minute me sikha deta hai; phir Kubernetes ki complexity samajhna aasan ho jata hai.

πŸ”œ Next Week

Week 5 me hum Compose ko pro level par le jayenge β€” multi-service apps, profiles, depends_on healthchecks, aur ek poora project: frontend + API + database, ek command me up!


Edit page
Share this post on:

Previous Post
Week 5: Containerizing a React Frontend
Next Post
Week 3: Persistent Storage Mastery