DevOps Classroom notes 10/Oct/2026

K8s Deployment Strategies:

• To Move or update one version to another version

1. Recreat:

• Kill all old pods at once and create new pods
• Will have down time

flowchart LR
    A[v1 v1 v1] --> B[all stopped - downtime] --> C[v2 v2 v2]

spec:
  Strategy:
     type: recreate

2. Rolling Update:

• it is default upgrade Strategie in k8s deployment
• pods will be replace gradually , a few min of time
• No downtime

flowchart LR
    S1[v1 v1 v1 v1] --> S2[v1 v1 v1 v2] --> S3[v1 v1 v2 v2] --> S4[v1 v2 v2 v2] --> S5[v2 v2 v2 v2]

spec:
  replicas: 4
  Strategy:
    type: RollingUpdate
    RollingUpdate:
      maxSurge: 1
      maxUnavailable: 0

3. Bule/green Deployment:

• Run Two full versions , switch traffic by changing the service selector

flowchart LR
    U[Users] --> SVC[Service<br/>selector: version=blue]
    SVC --> B[Blue v1 - live]
    G[Green v2 - idle/testing]
    SVC -. switch selector .-> G

4. Canary deployement:

• creating pods and load will be distubuted by %
• Traffic will divided by Ratio
• No downtime

flowchart LR
    U[Users] --> SVC[Service<br/>selector: app=myapp]
    SVC -->|~90%| S[Stable v1 x 9 Pods]
    SVC -->|~10%| C[Canary v2 x 1 Pod]

5. A/B Testing:

• it is similar as Canary deployment , based on request it will create group and random percentage to scale the new version

flowchart LR
    U[Request] --> R{Router<br/>header / cookie?}
    R -->|group A| A[v1]
    R -->|group B| B[v2]

6. Shadow:

• copy the traffic to new version
• safer to test work loads
• it is required service mesh or proxy configuration.

flowchart LR
    U[Users] --> P[Proxy / Mesh]
    P --> V1[v1 - responds to users]
    P -. mirrored copy .-> V2[v2 - response ignored]

Choosing the Right Strategy

Situation	Strategy
Dev/test, downtime acceptable	Recreate
Normal production release	Rolling Update
Need instant rollback, can afford 2x resources	Blue/Green
Risky change, want gradual exposure	Canary
Compare features by user group	A/B Testing
Validate under real traffic with zero risk	Shadow

---

Task:

1. create deployment using api referance Deployment API Offical docs
2. create new deployment Strategies like recreat , rolling update bule-green methods
3. try from apps , like jenkins , nop, apache2 , nginx

docker run -d -v jenkins_home:/var/jenkins_home -p 8080:8080 -p 50000:50000 --restart=on-failure jenkins/jenkins:lts-jdk21