Note_Tech

All technological notes.

---

Project maintained by simonangel-fong Hosted on GitHub Pages — Theme by mattgraham

Kubernetes: Deployment

Back

• Kubernetes: Deployment
  • Deployment
    • Deployment status
    • Additional Labels
    • Deployment vs ReplicaSet
    • Deployment Deletion
  • Imperative Command
  • Declarative Manifest
    • Key fields
  • Scaling Deployment
  • Good Practice: Omitting replicas field
    • Lab: omitting replicas field
      • Problem
      • Solution
  • Lab: Default Replica and strategy

---

Deployment

• Deployment

  • an abstraction layer over ReplicaSet
  • manage the deployment of ReplicaSet.
    • can update the ReplicaSet and roll back to the previous version.

• mainly used for stateless workloads

  • the pods created with deployment are fungible

• Major roles:

  • Declarative Updates:
    • declare the desired state with YAML (or CLI)
  • Scaling
    • Easily scale applications up or down
  • Rolling Updates
    • Deploy new versions of your app without downtime.
  • Rollback
    • If something goes wrong, you can roll back to a previous version
  • Self-healing
    • If a Pod crashes or a node goes down, the Deployment (via ReplicaSet) ensures new Pods are scheduled.

---

Deployment status

• Progressing;
  • Deployment creates a new ReplicaSet.
  • Deployment is scaling up its newest ReplicaSet.
  • Deployment is scaling down its older ReplicaSet(s).
  • New Pods become ready or available (ready for at least MinReadySeconds).
• complete:
  • All of the replicas have been updated to the latest version
  • All of the replicas are available.
  • No old replicas for the Deployment are running.

• Failed

  • Insufficient quota
  • Readiness probe failures
  • Image pull errors
  • Insufficient permissions
  • Limit ranges
  • Application runtime misconfiguration
• Troubleshooting Deployment failure:
  • kubectl describe:
    • get details of deploy and replicaset
    • event: show failure event
    • condition: show Type, Status, Reason
  • kubectl get -o yaml
    • output the detail message of deployment and replicaset

---

Additional Labels

• pod-template-hash Label
  • Automatically add to the underlying rs and pod
    • become part of the underlying RS name
  • calculated from the contents of the Pod template
• When the Pod template gets changed, a new ReplicaSet is created.

---

Deployment vs ReplicaSet

• ReplicaSet: Ensures a fixed number of Pods are running.

• Deployment: Manages ReplicaSets, and adds features like rolling updates and rollbacks.

  • Most of the time, you won’t create ReplicaSets directly — you use a Deployment.
• Deployment controls the ReplicaSet, which in turn controls the individual Pods.

• a ReplicaSet object that’s automatically created when creating the Deployment object.

  • A Deployment object doesn‘t directly manage the Pod objects, but manages them through a ReplicaSet object

• Updating Pod template

  • ReplicaSet: update only the new created pod
  • Deploymen: update pods immediately
• Existing pod: the existing pods with matching labels
  • ReplicaSet: controlled by RS
  • Deploymen: not controlled by Deploy, because additional label pod-template-hash will be added.

---

Deployment Deletion

• when you delete a Deployment object, the underlying ReplicaSet and Pods are also deleted.

• Preserving the ReplicaSet and Pods

  • kubectl delete deploy NAME --cascade=orphan

• Adopting an existing ReplicaSet and Pods

  • Recreate deployment if the pod template unchanged.
  • because the additional label pod-template-hash is calculated based on the pod template.

---

Imperative Command

• CRUD

Command	Description
kubectl explain deploy	show deployment documentation
kubectl get deploy	List all Deployments in the current namespace.
kubectl describe deploy deploy_name	Show detailed information about a specific Deployment.
kubectl create deploy deploy_name --image=img_name	Create a deployment using image
kubectl create deploy nginx --image=nginx --dry-run=client --replicas=4 -o yaml	Show the deployemnt in yaml file
kubectl set image deploy_name nginx=nginx:1.25	Update the container image in a deployment
kubectl delete deploy deploy_name	Delete a Deployment by name.
kubectl scale deploy deploy_name --replicas=count-num	Scale the number of replicas for a Deployment.

• List Deploy: kubectl get deploy
  • NAME:
    • Deployment name
    • nignx
  • READY:
    • e.g.,3/3
    • the number of running pod/the desired number of pods
  • UP-TO-DATE
    • e.g.,3
    • the number of Pods that runs the current version
    • used to indicate whether rollout is still in progress.
  • AVAILABLE
    • e.g.,3
    • the number of pods that both are ready and have been ready for spec.minReadySeconds
    • used for rolling updates and zero-downtime guarantees.
  • AGE
    • e.g., 18m
    • Time since the Deployment object was created
    • Get updated when
      • cluster restarted
      • rolled out a new version
      • Pod crashed and was recreated

---

Declarative Manifest

Key fields

• spec.replicas:
  • desired number of replica
• spec.selector:
  • label selector
  • matchLabels: a map of lables
  • matchExpressions: a list of label selector requirements
• spec.template:
  • Pod template
• spec.strategy:
  • how Pods are replaced when updating the Pod template.

---

Scaling Deployment

• When scaling a Deployment, the Deployment controller does nothing but scale the underlying ReplicaSet

  • Deployment controller updates the desired number of the replicas in the ReplicaSet
  • then the ReplicaSet controller to do the rest.

• If scale the underlying RS, the RS controller scales the number of pod imdiately.

  • then the deploy controller detects the difference between the desired replicas in the deploy and the underlying rs, and update the replicas in the rs
  • finally, rs controller reconcile the pod number.

• If the underlying rs object get deleted, the deploy controller will recreate the rs

---

Good Practice: Omitting replicas field

• Practical but problematical scenario:

  • define replicas=3 in manifest
  • scale to 5 with kubectl scale
  • update manifest without changing the replicas field and apply
    • replicas return to 3, disrupting the service.

• To avoid accidentally scaling a Deployment each time you reapply its manifest file

  • omitting the replicas field, which creates only 1 replica when creating
  • then kubectl scale to update the replica number
    • the replica number won’t be touched even reapply the manifest.

• if the replica already apply:

  • kubectl apply edit-last-applied deploy NAME: to remove the replicas field in the snapshot

---

Lab: omitting replicas field

Problem

• manifest

# demo-deploy-with-replica.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: demo-deploy-with-replica
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - image: nginx
          name: nginx

kubectl apply -f demo-deploy-with-replica.yaml
# deployment.apps/demo-deploy-with-replica created

kubectl get deploy
# NAME                       READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-with-replica   3/3     3            3           25s

• Manually scale

kubectl scale deploy demo-deploy-with-replica --replicas=5
# deployment.apps/demo-deploy-with-replica scaled

# confirm
kubectl get deploy
# NAME                       READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-with-replica   5/5     5            5           2m56s

• reapply the manifest

kubectl apply -f demo-deploy-with-replica.yaml
# deployment.apps/demo-deploy-with-replica configured

# confirm: replica back to 3
kubectl get deploy
# NAME                       READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-with-replica   3/3     3            3           3m53s

the replica number change back to 3

---

Solution

• manifest without replica

# demo-deploy-without-replica.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: demo-deploy-without-replica
spec:
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - image: nginx
          name: nginx

kubectl apply -f demo-deploy-without-replica.yaml
# deployment.apps/demo-deploy-without-replica created

kubectl get deploy
# NAME                          READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-without-replica   1/1     1            1           22s

• Manually scale

kubectl scale deploy demo-deploy-without-replica --replicas=5
# deployment.apps/demo-deploy-without-replica scaled

kubectl get deploy
# NAME                          READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-without-replica   5/5     5            5           2m33s

• make some change and reapply manifest
  • add label: tier=web

kubectl apply -f demo-deploy-without-replica.yaml
# deployment.apps/demo-deploy-without-replica configured

kubectl get deploy demo-deploy-without-replica
# NAME                          READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-without-replica   5/5     5            5           5m29s

the replica remain 5.

---

Lab: Default Replica and strategy

# demo-deploy-default-replica.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: demo-deploy-default-replica
spec:
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
        type: front-end
    spec:
      containers:
        - name: nginx
          image: nginx

# create deployment
kubectl create -f demo-deploy-default-replica.yaml
# deployment.apps/demo-deploy-default-replica created

kubectl get deploy
# NAME                          READY   UP-TO-DATE   AVAILABLE   AGE
# demo-deploy-default-replica   1/1     1            1           96s

# also create rs
kubectl get rs
# NAME                                     DESIRED   CURRENT   READY   AGE
# demo-deploy-default-replica-65cf588b89   1         1         1       106s

kubectl get pods
# NAME                                           READY   STATUS    RESTARTS   AGE
# demo-deploy-default-replica-65cf588b89-v4kpd   1/1     Running   0          2m20s

# confirm deploy:
#   default replica=1; default StrategyType=RollingUpdate; RollingUpdateStrategy:25% max unavailable, 25% max surge
kubectl describe deploy demo-deploy-default-replica
# Selector:               app=nginx
# Replicas:               1 desired | 1 updated | 1 total | 1 available | 0 unavailable
# StrategyType:           RollingUpdate
# RollingUpdateStrategy:  25% max unavailable, 25% max surge

# confirm rs:
#   Controlled by: deloy
#   additional label added: pod-template-hash + selector labels
kubectl describe rs demo-deploy-default-replica
# Labels:         app=nginx
#                 pod-template-hash=65cf588b89
#                 type=front-end
# Controlled By:  Deployment/demo-deploy-default-replica
# Replicas:       1 current / 1 desired

# confirm pod:
#   Controlled by rs
#   additional label added: pod-template-hash + selector labels
kubectl describe pod demo-deploy-default-replica
# Controlled By:  ReplicaSet/demo-deploy-default-replica-65cf588b89
# Labels:           app=nginx
#                   pod-template-hash=65cf588b89
#                   type=front-end

---