If your pods have no requirements for how or where to run, you can let Karpenter choose nodes from the full range of available cloud provider resources. However, by taking advantage of Karpenter’s model of layered constraints, you can be sure that the precise type and amount of resources needed are available to your pods. Reasons for constraining where your pods run could include:
- Needing to run in zones where dependent applications or storage are available
- Requiring certain kinds of processors or other hardware
- Wanting to use techniques like topology spread to help ensure high availability
Your Cloud Provider defines the first layer of constraints, including all instance types, architectures, zones, and purchase types available to its cloud. The cluster administrator adds the next layer of constraints by creating one or more provisioners. The final layer comes from you adding specifications to your Kubernetes pod deployments. Pod scheduling constraints must fall within a provisioner’s constraints or the pods will not deploy. For example, if the provisioner sets limits that allow only a particular zone to be used, and a pod asks for a different zone, it will not be scheduled.
Constraints you can request include:
- Resource requests: Request that certain amount of memory or CPU be available.
- Node selection: Choose to run on a node that is has a particular label (
- Node affinity: Draws a pod to run on nodes with particular attributes (affinity).
- Topology spread: Use topology spread to help ensure availability of the application.
- Pod affinity/anti-affinity: Draws pods towards or away from topology domains based on the scheduling of other pods.
Karpenter supports standard Kubernetes scheduling constraints. This allows you to define a single set of rules that apply to both existing and provisioned capacity.
NoteKarpenter supports specific Well-Known Labels, Annotations and Taints that are useful for scheduling.
Within a Pod spec, you can both make requests and set limits on resources a pod needs, such as CPU and memory. For example:
apiVersion: v1 kind: Pod metadata: name: myapp spec: containers: - name: app image: myimage resources: requests: memory: "128Mi" cpu: "500m" limits: memory: "256Mi" cpu: "1000m"
In this example, the container is requesting 128MiB of memory and .5 CPU.
Its limits are set to 256MiB of memory and 1 CPU.
Instance type selection math only uses
limits may be configured to enable resource oversubscription.
See Managing Resources for Containers for details on resource types supported by Kubernetes, Specify a memory request and a memory limit for examples of memory requests, and Provisioner Configuration for a list of supported resources.
Accelerator (e.g., GPU) values include
Karpenter supports accelerators, such as GPUs.
Additionally, include a resource requirement in the workload manifest. This will cause the GPU dependent pod to be scheduled onto the appropriate node.
Here is an example of an accelerator resource in a workload manifest (e.g., pod):
spec: template: spec: containers: - resources: limits: nvidia.com/gpu: "1"
If you are provisioning GPU nodes, you need to deploy an appropriate GPU device plugin daemonset for those nodes. Without the daemonset running, Karpenter will not see those nodes as initialized. Refer to general Kubernetes GPU docs and the following specific GPU docs:
Pod ENI Resources (Security Groups for Pods)
Pod ENI is a feature of the AWS VPC CNI Plugin which allows an Elastic Network Interface (ENI) to be allocated directly to a Pod. When enabled, the
vpc.amazonaws.com/pod-eni extended resource is added to supported nodes. The Pod ENI feature can be used independently, but is most often used in conjunction with Security Groups for Pods. Follow the below instructions to enable support for Pod ENI and/or Security Groups for Pods in Karpenter.
NoteYou must enable Pod ENI support in the AWS VPC CNI Plugin before enabling Pod ENI support in Karpenter. Please refer to the Security Groups for Pods documentation for instructions.
Now that Pod ENI support is enabled in the AWS VPC CNI Plugin, you can enable Pod ENI support in Karpenter by setting the
settings.aws.enablePodENI Helm chart value to
Here is an example of a pod-eni resource defined in a deployment manifest:
spec: template: spec: containers: - resources: limits: vpc.amazonaws.com/pod-eni: "1"
nodeSelector you can ask for a node that matches selected key-value pairs.
This can include well-known labels or custom labels you create yourself.
You can use
affinity to define more complicated constraints, see Node Affinity for the complete specification.
Well-known labels may be specified as provisioner requirements or pod scheduling constraints. You can also define your own custom labels by specifying
labels on your Provisioner and select them using
nodeSelectors on your Pods.
WarningTake care to ensure the label domains are correct. A well known label like
karpenter.k8s.aws/instance-familywill enforce node properties, but may be confused with
node.kubernetes.io/instance-family, which is unknown to Karpenter, and treated as a custom label which will not enforce node properties.
|topology.kubernetes.io/zone||us-east-2a||Zones are defined by your cloud provider (aws)|
|node.kubernetes.io/instance-type||g4dn.8xlarge||Instance types are defined by your cloud provider (aws)|
|kubernetes.io/os||linux||Operating systems are defined by GOOS values on the instance|
|kubernetes.io/arch||amd64||Architectures are defined by GOARCH values on the instance|
|karpenter.sh/capacity-type||spot||Capacity types include
|karpenter.k8s.aws/instance-hypervisor||nitro||[AWS Specific] Instance types that use a specific hypervisor|
|karpenter.k8s.aws/instance-category||g||[AWS Specific] Instance types of the same category, usually the string before the generation number|
|karpenter.k8s.aws/instance-generation||4||[AWS Specific] Instance type generation number within an instance category|
|karpenter.k8s.aws/instance-family||g4dn||[AWS Specific] Instance types of similar properties but different resource quantities|
|karpenter.k8s.aws/instance-size||8xlarge||[AWS Specific] Instance types of similar resource quantities but different properties|
|karpenter.k8s.aws/instance-cpu||32||[AWS Specific] Number of CPUs on the instance|
|karpenter.k8s.aws/instance-memory||131072||[AWS Specific] Number of mebibytes of memory on the instance|
|karpenter.k8s.aws/instance-pods||110||[AWS Specific] Number of pods the instance supports|
|karpenter.k8s.aws/instance-gpu-name||t4||[AWS Specific] Name of the GPU on the instance, if available|
|karpenter.k8s.aws/instance-gpu-manufacturer||nvidia||[AWS Specific] Name of the GPU manufacturer|
|karpenter.k8s.aws/instance-gpu-count||1||[AWS Specific] Number of GPUs on the instance|
|karpenter.k8s.aws/instance-gpu-memory||16384||[AWS Specific] Number of mebibytes of memory on the GPU|
|karpenter.k8s.aws/instance-local-nvme||900||[AWS Specific] Number of gibibytes of local nvme storage on the instance|
Karpenter is aware of several well-known labels, deriving them from instance type details. If you specify a
nodeSelector or a required
nodeAffinity using a label that is not well-known to Karpenter, it will not launch nodes with these labels and pods will remain pending. For Karpenter to become aware that it can schedule for these labels, you must specify the label in the Provisioner requirements with the
requirements: - key: user.defined.label/type operator: Exists
Here is an example of a
nodeSelector for selecting nodes:
nodeSelector: topology.kubernetes.io/zone: us-west-2a karpenter.sh/capacity-type: spot
This example features a well-known label (
topology.kubernetes.io/zone) and a label that is well known to Karpenter (
If you want to create a custom label, you should do that at the provisioner level. Then the pod can declare that custom label.
See nodeSelector in the Kubernetes documentation for details.
Examples below illustrate how to use Node affinity to include (
In) and exclude (
See Node affinity for details.
When setting rules, the following Node affinity types define how hard or soft each rule is:
- requiredDuringSchedulingIgnoredDuringExecution: This is a hard rule that must be met.
- preferredDuringSchedulingIgnoredDuringExecution: This is a preference, but the pod can run on a node where it is not guaranteed.
IgnoredDuringExecution part of each tells the pod to keep running, even if conditions change on the node so the rules no longer matched.
You can think of these concepts as
preferred, since Kubernetes never implemented other variants of these rules.
All examples below assume that the provisioner doesn’t have constraints to prevent those zones from being used.
The first constraint says you could use
us-west-2b, the second constraint makes it so only
us-west-2b can be used.
affinity: nodeAffinity: requiredDuringSchedulingIgnoredDuringExecution: nodeSelectorTerms: - matchExpressions: - key: "topology.kubernetes.io/zone" operator: "In" values: ["us-west-2a, us-west-2b"] - key: "topology.kubernetes.io/zone" operator: "In" values: ["us-west-2b"]
Changing the second operator to
NotIn would allow the pod to run in
- key: "topology.kubernetes.io/zone" operator: "In" values: ["us-west-2a, us-west-2b"] - key: "topology.kubernetes.io/zone" operator: "NotIn" values: ["us-west-2b"]
Continuing to add to the example,
nodeAffinity lets you define terms so if one term doesn’t work it goes to the next one.
us-west-2a is not available, the second term will cause the pod to run on a spot instance in
affinity: nodeAffinity: requiredDuringSchedulingIgnoredDuringExecution: nodeSelectorTerms: - matchExpressions: # OR - key: "topology.kubernetes.io/zone" # AND operator: "In" values: ["us-west-2a, us-west-2b"] - key: "topology.kubernetes.io/zone" # AND operator: "NotIn" values: ["us-west-2b"] - matchExpressions: # OR - key: "karpenter.sh/capacity-type" # AND operator: "In" values: ["spot"] - key: "topology.kubernetes.io/zone" # AND operator: "In" values: ["us-west-2d"]
In general, Karpenter will go through each of the
nodeSelectorTerms in order and take the first one that works.
However, if Karpenter fails to provision on the first
nodeSelectorTerms, it will try again using the second one.
If they all fail, Karpenter will fail to provision the pod.
Karpenter will backoff and retry over time.
So if capacity becomes available, it will schedule the pod without user intervention.
Taints and tolerations
Taints are the opposite of affinity. Setting a taint on a node tells the scheduler to not run a pod on it unless the pod has explicitly said it can tolerate that taint. This example shows a Provisioner that was set up with a taint for only running pods that require a GPU, such as the following:
apiVersion: karpenter.sh/v1alpha5 kind: Provisioner metadata: name: gpu spec: requirements: - key: karpenter.k8s.aws/instance-family operator: In values: - p3 taints: - key: nvidia.com/gpu value: true effect: "NoSchedule"
For a pod to request to run on a node that has provisioner, it could set a toleration as follows:
apiVersion: v1 kind: Pod metadata: name: mygpupod spec: containers: - name: gpuapp resources: requests: nvidia.com/gpu: 1 limits: nvidia.com/gpu: 1 image: mygpucontainer tolerations: - key: "nvidia.com/gpu" operator: "Exists" effect: "NoSchedule"
See Taints and Tolerations in the Kubernetes documentation for details.
By using the Kubernetes
topologySpreadConstraints you can ask the provisioner to have pods push away from each other to limit the blast radius of an outage.
Think of it as the Kubernetes evolution for pod affinity: it lets you relate pods with respect to nodes while still allowing spread.
spec: topologySpreadConstraints: - maxSkew: 1 topologyKey: "topology.kubernetes.io/zone" whenUnsatisfiable: ScheduleAnyway labelSelector: matchLabels: dev: jjones - maxSkew: 1 topologyKey: "kubernetes.io/hostname" whenUnsatisfiable: ScheduleAnyway labelSelector: matchLabels: dev: jjones - maxSkew: 1 topologyKey: "karpenter.sh/capacity-type" whenUnsatisfiable: ScheduleAnyway labelSelector: matchLabels: dev: jjones
Adding this to your podspec would result in:
- Pods being spread across zones, hosts, and capacity-type (
labelSelectorwill include all pods with the label of
dev=jjonesin topology calculations. It is recommended to use a selector to match all pods in a deployment.
- No more than one pod difference in the number of pods on each host (
maxSkew). For example, if there were three nodes and five pods the pods could be spread 1, 2, 2 or 2, 1, 2 and so on. If instead the spread were 5, pods could be 5, 0, 0 or 3, 2, 0, or 2, 1, 2 and so on.
The three supported
topologyKey values that Karpenter supports are:
See Pod Topology Spread Constraints for details.
By using the
podAntiAffinity configuration on a pod spec, you can inform the provisioner of your desire for pods to schedule together or apart with respect to different topology domains. For example:
spec: affinity: podAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchExpressions: - key: system operator: In values: - backend topologyKey: topology.kubernetes.io/zone podAntiAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchLabels: app: inflate topologyKey: kubernetes.io/hostname
The above pod affinity rule would cause the pod to only schedule in zones where a pod with the label
system=backend is already running.
The anti-affinity rule would cause it to avoid running on any node with a pod labeled
app=inflate. If this anti-affinity term was on a deployment pod spec along with a matching
app=inflate label, it would prevent more than one pod from the deployment from running on any single node.
See Inter-pod affinity and anti-affinity in the Kubernetes documentation for details.
Persistent Volume Topology
Karpenter automatically detects storage scheduling requirements and includes them in node launch decisions.
In the following example, the
StorageClass defines zonal topologies for
us-west-2b and binding mode
When the pod is created, Karpenter follows references from the
StorageClass and identifies that this pod requires storage in
It randomly selects
us-west-2a, provisions a node in that zone, and waits for kube-scheduler to bind the pod to the node.
The CSI driver creates a
PersistentVolume according to the
PersistentVolumeClaim and gives it a node affinity rule for
Later on, the pod is deleted and a new pod is created that requests the same claim. This time, Karpenter identifies that a
PersistentVolume already exists for the
PersistentVolumeClaim, and includes its zone
us-west-2a in the pod’s scheduling requirements.
apiVersion: v1 kind: Pod metadata: name: app spec: containers: ... volumes: - name: storage persistentVolumeClaim: claimName: ebs-claim --- kind: StorageClass apiVersion: storage.k8s.io/v1 metadata: name: ebs provisioner: ebs.csi.aws.com volumeBindingMode: WaitForFirstConsumer allowedTopologies: - matchLabelExpressions: - key: topology.ebs.csi.aws.com/zone values: ["us-west-2a", "us-west-2b"] --- apiVersion: v1 kind: PersistentVolumeClaim metadata: name: ebs-claim spec: accessModes: - ReadWriteOnce storageClassName: ebs resources: requests: storage: 4Gi
☁️ AWS Specific
The EBS CSI driver uses
topology.ebs.csi.aws.com/zone instead of the standard
topology.kubernetes.io/zone label. Karpenter is aware of label aliasing and translates this label into
topology.kubernetes.io/zone in memory. When configuring a
StorageClass for the EBS CSI Driver, you must use
NoteThe topology key
topology.kubernetes.io/regionis not supported. Legacy in-tree CSI providers specify this label. Instead, install an out-of-tree CSI provider. Learn more about moving to CSI providers.
Karpenter allows you to order your provisioners using the
.spec.weight field so that the node scheduler will deterministically attempt to schedule with one provisioner before another. Below are a few example use-cases that are now supported with the provisioner weighting semantic.
Savings Plans and Reserved Instances
To enable this, you will need to tell the Karpenter controllers which instance types to prioritize and what is the maximum amount of capacity that should be provisioned using those instance types. We can set the
.spec.limits on the provisioner to limit the capacity that can be launched by this provisioner. Combined with the
.spec.weight value, we can tell Karpenter to pull from instance types in the reserved provisioner before defaulting to generic instance types.
Reserved Instance Provisioner
apiVersion: karpenter.sh/v1alpha5 kind: Provisioner metadata: name: reserved-instance spec: weight: 50 requirements: - key: "node.kubernetes.io/instance-type" operator: In values: ["c4.large"] limits: cpu: 100
apiVersion: karpenter.sh/v1alpha5 kind: Provisioner metadata: name: default spec: requirements: - key: karpenter.sh/capacity-type operator: In values: ["spot", "on-demand"] - key: kubernetes.io/arch operator: In values: ["amd64"]
Default Node Configuration
Pods that do not specify node selectors or affinities can potentially be assigned to any node with any configuration. There may be cases where you require these pods to schedule to a specific capacity type or architecture but assigning the relevant node selectors or affinities to all these workload pods may be too tedious or infeasible. Instead, we want to define a cluster-wide default configuration for nodes launched using Karpenter.
By assigning a higher
.spec.weight value and restricting a provisioner to a specific capacity type or architecture, we can set default configuration for the nodes launched by pods that don’t have node configuration restrictions.
apiVersion: karpenter.sh/v1alpha5 kind: Provisioner metadata: name: default spec: weight: 50 requirements: - key: karpenter.sh/capacity-type operator: In values: ["spot", "on-demand"] - key: kubernetes.io/arch operator: In values: ["amd64"]
ARM-64 Specific Provisioner
apiVersion: karpenter.sh/v1alpha5 kind: Provisioner metadata: name: arm64-specific spec: requirements: - key: karpenter.sh/capacity-type operator: In values: ["spot", "on-demand"] - key: kubernetes.io/arch operator: In values: ["arm64"] - key: node.kubernetes.io/instance-type operator: In values: ["a1.large", "a1.xlarge"]
NoteBased on the way that Karpenter performs pod batching and bin packing, it is not guaranteed that Karpenter will always choose the highest priority provisioner given specific requirements. For example, if a pod can’t be scheduled with the highest priority provisioner it will force creation of a node using a lower priority provisioner which may allow other pods from that batch to also schedule on that node. The behavior may also occur if existing capacity is available, as the kube-scheduler will schedule the pods instead of allowing Karpenter to provision a new node.
Advanced Scheduling Techniques
Exists operator can be used on a provisioner to provide workload segregation across nodes.
... requirements: - key: company.com/team operator: Exists ...
With the requirement on the provisioner in place, workloads can optionally specify a custom value as a required node affinity or node selector. Karpenter will then label the nodes it launches for these pods which prevents
kube-scheduler from scheduling conflicting pods to those nodes. This provides a way to more dynamically isolate workloads without requiring a unique provisioner for each workload subset.
nodeSelector: company.com/team: team-a
NoteIf a workload matches the provisioner but doesn’t specify a label, Karpenter will generate a random label for the node.
On-Demand/Spot Ratio Split
Taking advantage of Karpenter’s ability to assign labels to node and using a topology spread across those labels enables a crude method for splitting a workload across on-demand and spot instances in a desired ratio.
To do this, we create a provisioner each for spot and on-demand with disjoint values for a unique new label called
capacity-spread. In the example below, we provide four unique values for the spot provisioner and one value for the on-demand provisioner. When we spread across our new label evenly, we’ll end up with a ratio of 4:1 spot to on-demand nodes.
WarningThis is not identical to a topology spread with a specified ratio. We are constructing ‘virtual domains’ to spread evenly across and the ratio of those ‘virtual domains’ to spot and on-demand happen to coincide with the desired spot to on-demand ratio. As an example, if you launch pods using the provided example, Karpenter will launch nodes with
capacity-spreadlabels of 1, 2, 3, 4, and 5.
kube-schedulerwill then schedule evenly across those nodes to give the desired ratio.
requirements: - key: "karpenter.sh/capacity-type" operator: In values: [ "spot"] - key: capacity-spread operator: In values: - "2" - "3" - "4" - "5"
requirements: - key: "karpenter.sh/capacity-type" operator: In values: [ "on-demand"] - key: capacity-spread operator: In values: - "1"
Workload Topology Spread Constraint
topologySpreadConstraints: - maxSkew: 1 topologyKey: capacity-spread whenUnsatisfiable: DoNotSchedule