Skip to main content

k8s

The k8s scope controls the deployment and execution of workflow applications in a Kubernetes cluster.

k8s.autoMountHostPaths​

Automatically mount host paths into the task pods (default: false). Only intended for development purposes when using a single node.

k8s.cleanup​

When true, successful pods are automatically deleted (default: true).

k8s.client​

Map of options for the Kubernetes HTTP client.

If this option is specified, it will be used instead of .kube/config.

Available options:

  • server
  • token
  • tokenFile
  • verifySsl
  • sslCert
  • sslCertFile
  • clientCert
  • clientCertFile
  • clientKey
  • clientKeyFile
k8s.clientRefreshInterval​
Added in version 26.04

The interval after which the Kubernetes client configuration is refreshed (default: 50.m).

This setting is useful when the Kubernetes authentication token has a limited lifespan and needs to be periodically refreshed. The client configuration will be automatically reloaded after the specified interval, allowing Nextflow to obtain fresh credentials from the Kubernetes configuration.

k8s.computeResourceType​
Changed in version 26.10

The default value was changed from Pod to Job.

Whether to use Kubernetes Pod or Job resource type to carry out Nextflow tasks (default: Job).

k8s.context​

The Kubernetes configuration context to use.

k8s.cpuLimits​
Added in version 24.04

When true, set both the pod CPUs request and limit to the value specified by the cpus directive, otherwise set only the request (default: false). This setting is useful when a K8s cluster requires a CPU limit to be defined through a LimitRange.

k8s.fetchNodeName​

Include the hostname of each task in the execution trace (default: false).

k8s.fuseDevicePlugin​
Added in version 24.04

The FUSE device plugin to be used when enabling Fusion in unprivileged mode (default: ['nextflow.io/fuse': 1]).

k8s.httpConnectTimeout​

The Kubernetes HTTP client request connection timeout e.g. 60.s.

k8s.httpReadTimeout​

The Kubernetes HTTP client request connection read timeout e.g. 60.s.

k8s.imagePullPolicy​

The strategy for pulling container images. Can be IfNotPresent, Always, Never.

k8s.launchDir​

The path where the workflow is launched and the user data is stored (default: <volume-claim-mount-path>/<user-name>). Must be a path in a shared K8s persistent volume.

k8s.namespace​

The Kubernetes namespace to use (default: default).

k8s.pod​

Additional pod configuration options such as environment variables, config maps, secrets, etc. Allows the same settings as the pod process directive.

k8s.projectDir​

The path where Nextflow projects are downloaded (default: <volume-claim-mount-path>/projects). Must be a path in a shared K8s persistent volume.

k8s.runAsUser​

The user ID to be used to run the containers. Shortcut for the securityContext option.

k8s.securityContext​

The security context to use for all pods.

k8s.serviceAccount​

The Kubernetes service account name to use.

k8s.storageClaimName​

The name of the persistent volume claim where the shared work directory is stored.

k8s.storageMountPath​

The mount path for the persistent volume claim (default: /workspace).

k8s.storageSubPath​

The path in the persistent volume to be mounted (default: /).

k8s.workDir​

The path of the shared work directory (default: <user-dir>/work). Must be a path in a shared K8s persistent volume.

k8s.debug.yaml​

Save the pod spec for each task to .command.yaml in the task directory (default: false).

k8s.retryPolicy.delay​

Delay when retrying failed API requests (default: 250.ms).

k8s.retryPolicy.jitter​

Jitter value when retrying failed API requests (default: 0.25).

k8s.retryPolicy.maxAttempts​

Max attempts when retrying failed API requests (default: 4).

k8s.retryPolicy.maxDelay​

Max delay when retrying failed API requests (default: 90.s).

On this Page