Gestisci l'elaborazione per i AI/ML carichi di lavoro con EKS Auto Mode e Karpenter

Per l'elenco completo, consulta le etichette supportate dalla modalità automatica EKS.

Per l'elenco completo, vedi Karpenter Labels Well-Known .

L'esempio seguente mostra una capacità statica NodePool che utilizza la modalità automatica EKS predefinita NodeClass e crea una capacità statica NodePool con 4 nodi (replicas) che possono essere al massimo 6 nodi (). limits.nodes

apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-static-inference
spec:
  replicas: 4
  template:
    spec:
      nodeClassRef:
        group: eks.amazonaws.com
        kind: NodeClass
        name: default
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["on-demand"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["g6e"]
        - key: "topology.kubernetes.io/zone"
          operator: In
          values: ["us-east-1a"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule
  limits:
    nodes: 6   # Allow temporary headroom during node replacement

Con Karpenter autogestito, la capacità statica è limitata dalla StaticCapacity funzione alpha (lanciata nella versione Karpenter v1.8), che deve essere abilitata nei valori Helm:

settings:
  featureGates:
    staticCapacity: true

Fa NodePool riferimento a un EC2NodeClass nome personalizzato my-nodeclass e crea uno statico NodePool con 4 nodi (replicas) che possono essere al massimo 6 nodi (). limits.nodes

apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-static-inference
spec:
  replicas: 4
  template:
    spec:
      nodeClassRef:
        group: karpenter.k8s.aws
        kind: EC2NodeClass
        name: my-nodeclass
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["on-demand"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["g6e"]
        - key: "topology.kubernetes.io/zone"
          operator: In
          values: ["us-east-1a"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule
  limits:
    nodes: 6   # Allow temporary headroom during node replacement

Creane una NodeClass che faccia riferimento alla tua prenotazione Capacity Block, quindi creane una NodePool che la utilizzi.

Con consolidateAfter: Never set, Karpenter non tenterà di sostituire, unire o terminare i nodi per ridurre i costi o imballare i carichi di lavoro in modo più efficiente. Questa opzione è consigliata per i Capacity Blocks perché la capacità è già prepagata.

apiVersion: eks.amazonaws.com/v1
kind: NodeClass
metadata:
  name: capacity-block-gpu
spec:
  capacityReservationSelectorTerms:
    - id: "cr-0123456789abcdef0"   # Your Capacity Block reservation ID
    # Alternative: select by tags
    # - tags:
    #     role: "production-inference"
    #   owner: "012345678901"
---
apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-capacity-block
spec:
  disruption:
    consolidationPolicy: WhenEmpty
    consolidateAfter: Never
  template:
    spec:
      nodeClassRef:
        group: eks.amazonaws.com
        kind: NodeClass
        name: capacity-block-gpu
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["reserved"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["p5", "p5e", "p5en", "p4d"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule

Creane uno EC2NodeClass che includa selettori AMI, sottorete e gruppi di sicurezza oltre acapacityReservationSelectorTerms, quindi creane uno NodePool che lo utilizzi.

Con consolidateAfter: Never set, Karpenter non tenterà di sostituire, unire o terminare i nodi per ridurre i costi o imballare i carichi di lavoro in modo più efficiente. Questa opzione è consigliata per i Capacity Blocks perché la capacità è già prepagata.

apiVersion: karpenter.k8s.aws/v1
kind: EC2NodeClass
metadata:
  name: capacity-block-gpu
spec:
  amiSelectorTerms:
    - alias: al2023@latest
  subnetSelectorTerms:
    - tags:
        karpenter.sh/discovery: ml-cluster
  securityGroupSelectorTerms:
    - tags:
        karpenter.sh/discovery: ml-cluster
  capacityReservationSelectorTerms:
    - id: "cr-0123456789abcdef0" # Your Capacity Block reservation ID
    # Alternative: select by tags
    # - tags:
    #     role: "production-inference"
    #   owner: "012345678901"
---
apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-capacity-block
spec:
  disruption:
    consolidationPolicy: WhenEmpty
    consolidateAfter: Never
  template:
    spec:
      nodeClassRef:
        group: karpenter.k8s.aws
        kind: EC2NodeClass
        name: capacity-block-gpu
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["reserved"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["p5", "p5e", "p5en", "p4d"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule

Crea un NodeClass with capacityReservationSelectorTerms e un NodePool che dia priorità reserved con fallback. on-demand Aggiungi topology.kubernetes.io/zone alla AZ dell'ODCR:

apiVersion: eks.amazonaws.com/v1
kind: NodeClass
metadata:
  name: odcr-gpu-production
spec:
  capacityReservationSelectorTerms:
    - id: "cr-0987654321fedcba0"
    # Alternative: select by tags
    # - tags:
    #     Purpose: "production-inference"
    #   owner: "012345678901"
---
apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-reserved-production
spec:
  disruption:
    consolidationPolicy: WhenEmpty
    consolidateAfter: Never
  template:
    spec:
      nodeClassRef:
        group: eks.amazonaws.com
        kind: NodeClass
        name: odcr-gpu-production
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["reserved", "on-demand"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["p5", "g6e"]
        - key: "topology.kubernetes.io/zone"
          operator: In
          values: ["us-east-1a"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule

Crea un selettore EC2NodeClass con AMI, sottorete e gruppo di sicurezza oltre acapacityReservationSelectorTerms, quindi crea: NodePool

apiVersion: karpenter.k8s.aws/v1
kind: EC2NodeClass
metadata:
  name: odcr-gpu-production
spec:
  amiSelectorTerms:
    - alias: al2023@latest
  subnetSelectorTerms:
    - tags:
        karpenter.sh/discovery: ml-cluster
  securityGroupSelectorTerms:
    - tags:
        karpenter.sh/discovery: ml-cluster
  capacityReservationSelectorTerms:
    - id: "cr-0987654321fedcba0"
---
apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-reserved-production
spec:
  disruption:
    consolidationPolicy: WhenEmpty
    consolidateAfter: Never
  template:
    spec:
      nodeClassRef:
        group: karpenter.k8s.aws
        kind: EC2NodeClass
        name: odcr-gpu-production
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["reserved", "on-demand"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["p5", "g6e"]
        - key: "topology.kubernetes.io/zone"
          operator: In
          values: ["us-east-1a"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule

apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-ondemand
spec:
  template:
    spec:
      nodeClassRef:
        group: eks.amazonaws.com
        kind: NodeClass
        name: default
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["on-demand"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["g6", "g6e", "g7e"]
        - key: "karpenter.k8s.aws/instance-gpu-manufacturer"
          operator: In
          values: ["nvidia"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule

apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-ondemand
spec:
  template:
    spec:
      nodeClassRef:
        group: karpenter.k8s.aws
        kind: EC2NodeClass
        name: default
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["on-demand"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["g6", "g6e", "g7e"]
        - key: "karpenter.k8s.aws/instance-gpu-manufacturer"
          operator: In
          values: ["nvidia"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule

EKS Auto Mode gestisce le interruzioni Spot in modo nativo. Non è richiesta alcuna coda SQS o Node Termination Handler.

apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-spot
spec:
  disruption:
    budgets:
      - nodes: 10%
    consolidationPolicy: WhenEmpty
    consolidateAfter: 1h
  template:
    spec:
      nodeClassRef:
        group: eks.amazonaws.com
        kind: NodeClass
        name: default
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["spot"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["g6", "g6e", "g7e"]
        - key: "karpenter.k8s.aws/instance-gpu-manufacturer"
          operator: In
          values: ["nvidia"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule
  limits:
    resources:
      nvidia.com/gpu: "64"

Self-managed Karpenter richiede di abilitare la gestione nativa delle interruzioni sul controller Karpenter (non su NodePool) configurando una coda di interruzione: una coda SQS che riceve gli eventi di interruzione Spot e Rebalance Recommendation di EC2. È possibile configurarla una sola volta al momento dell'installazione.

Se installi Karpenter direttamente con Helm, imposta settings.interruptionQueue il tuo: values.yaml

# karpenter values.yaml (Helm)
settings:
  clusterName: my-cluster
  interruptionQueue: my-queue   # Name of the SQS queue receiving Spot events

Se avviate Karpenter coneksctl, withSpotInterruptionQueue: true impostatelo nel file di configurazione del cluster. eksctlcrea la coda e le EventBridge regole SQS e configura il controller Karpenter per utilizzarle.

# eksctl ClusterConfig
karpenter:
  version: "${KARPENTER_VERSION}"
  withSpotInterruptionQueue: true

Una volta che il controller è configurato per utilizzare la coda, non è necessaria alcuna configurazione aggiuntiva per le singole risorse. NodePool La gestione delle interruzioni si applica a livello di cluster.

apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: gpu-spot
spec:
  disruption:
    budgets:
      - nodes: 10%
    consolidationPolicy: WhenEmpty
    consolidateAfter: 1h
  template:
    spec:
      nodeClassRef:
        group: karpenter.k8s.aws
        kind: EC2NodeClass
        name: default
      requirements:
        - key: "karpenter.sh/capacity-type"
          operator: In
          values: ["spot"]
        - key: "karpenter.k8s.aws/instance-family"
          operator: In
          values: ["g6", "g6e", "g7e"]
        - key: "karpenter.k8s.aws/instance-gpu-manufacturer"
          operator: In
          values: ["nvidia"]
      taints:
        - key: nvidia.com/gpu
          value: "true"
          effect: NoSchedule
  limits:
    resources:
      nvidia.com/gpu: "64"