[애플리케이션 스케줄링과 라이프사이클 관리] 테인트를 활용한 스케줄링 이해

Taint와 Toleration 원리

• 테인트는 "오점을 남기다, 더럽히다"라는 뜻을 가짐
• 테인트는 노드에 설정하며 설정된 노드에는 "어떤" 오점을 남김
• 톨러레이션은 "용인하다, 견디다"란 뜻을 가짐
• 노드가 갖고 있는 오점을 용인하는 옵션을 포드에 설정 -> 노드에 배치 가능
• 톨러레이션 옵션은 노드 셀렉터 등과 다르게 아무 오점이 없어도 배치가 될 수 있으며, 만약 오점이 있는 경우에는 톨러레이션 세팅이 필수적으로 필요

Taint 현황 확인하기

• kubectl describe node <노드 이름> | grep -i Taints
• kubectl get nodes -o json | jq .items[].spec.taints

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl describe node gke-artbridge-default-pool-65403ed8-fz5f | grep -i taints
                    node.gke.io/last-applied-node-taints:
Taints:             <none>

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl get nodes -o json | jq .items[].spec.taints
null
null
null

노드에 테인트 추가하기

• 사용자가 원하는 노드에 테인트 설정 가능
• kubectl taint nodes <노드이름> key1=value1:NoExecute

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl get pods -A -o wide | grep gke-artbridge-default-pool-65403ed8-7zvx
argocd        argocd-application-controller-0                       1/1     Running   0             5m32s   10.40.0.175   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
argocd        argocd-applicationset-controller-5dffff55bd-pzpv6     1/1     Running   0             5m34s   10.40.0.172   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
argocd        argocd-dex-server-656864dd94-ph98s                    1/1     Running   0             11d     10.40.0.7     gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
argocd        argocd-notifications-controller-567f8cdddc-cxg7s      1/1     Running   0             5m34s   10.40.0.171   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
argocd        argocd-redis-b5d6bf5f5-mbfck                          1/1     Running   0             11d     10.40.0.8     gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
argocd        argocd-repo-server-7555f4b465-g9btp                   1/1     Running   0             5m34s   10.40.0.173   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
argocd        argocd-server-7f758fccf6-mpds7                        1/1     Running   0             11d     10.40.0.9     gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
default       config-server                                         1/1     Running   0             5d20h   10.40.0.168   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
default       exhibition-sonar-deployment-57b47c487d-5tt9m          1/1     Running   0             2d15h   10.40.0.169   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
default       http-go-r2bbv                                         1/1     Running   0             39s     10.40.0.177   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
default       mariadb-5bfcbc8dd5-8w8n8                              1/1     Running   0             5m33s   10.40.0.176   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
default       nginx-sidecar                                         3/3     Running   0             8d      10.40.0.22    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
flask0        flask-6b7fbcfd94-h6xd5                                1/1     Running   0             11d     10.40.0.10    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
gmp-system    collector-5mkmp                                       2/2     Running   0             10d     10.40.0.17    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   fluentbit-gke-p5n7f                                   2/2     Running   0             12d     10.128.0.3    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   gke-metrics-agent-hhbz2                               2/2     Running   0             12d     10.128.0.3    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   kube-proxy-gke-artbridge-default-pool-65403ed8-7zvx   1/1     Running   0             12d     10.128.0.3    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   pdcsi-node-l9dq5                                      2/2     Running   0             10d     10.128.0.3    gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
mychart2      mychart2-test-7699f994d8-jdrzb                        1/1     Running   0             5m34s   10.40.0.174   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>

---

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl taint nodes gke-artbridge-default-pool-65403ed8-7zvx key1=value1:NoExecute
node/gke-artbridge-default-pool-65403ed8-7zvx tainted

---

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl get pods -A -o wide | grep gke-artbridge-default-pool-65403ed8-7zvx
gmp-system    collector-5mkmp                                       2/2     Running             0             10d     10.40.0.17   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   fluentbit-gke-p5n7f                                   2/2     Running             0             12d     10.128.0.3   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   gke-metrics-agent-hhbz2                               2/2     Running             0             12d     10.128.0.3   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   kube-proxy-gke-artbridge-default-pool-65403ed8-7zvx   1/1     Running             0             12d     10.128.0.3   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>
kube-system   pdcsi-node-l9dq5                                      2/2     Running             0             10d     10.128.0.3   gke-artbridge-default-pool-65403ed8-7zvx   <none>           <none>

---

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl describe pod kube-proxy-gke-artbridge-default-pool-65403ed8-7zvx -n kube-system | grep -A7 Tolerations
Tolerations:       :NoExecute op=Exists
                   :NoSchedule op=Exists
Events:            <none>

• NoExecute op=Exists: "NoExecute" Taint에 대해 무시
• NoSchedule op=Exists: "NoSchedule" Taint에 대해 무시

---

원상 복구

imkunyoung@cloudshell:~ (kubernetes-397511)$ kubectl taint nodes gke-artbridge-default-pool-65403ed8-7zvx key1=value1:NoExecute-
node/gke-artbridge-default-pool-65403ed8-7zvx untainted

포드에 톨러레이션 추가

# pod-with-toleration.yaml
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:
  - name: nginx
    image: nginx
    imagePullPolicy: IfNotPresent
  tolerations:
  - key: "key1"
    operator: "Exists"
    value: "value1"
    effect: "NoExecute"

Effect 종류	설명
NoSchedule	해당 노드에 포드를 스케줄링하지 않음
PreferNoSchedule	가능한 경우 해당 노드에 포드를 스케줄링하지 않도록 함
NoExecute	해당 노드에서 포드를 실행 중인 경우 포드를 중단시킴