Amazon EKS — FSx CSI Driver

1 minute read

Amazon FSx provides high-performance file systems purpose-built for specific workloads. The FSx CSI drivers let you mount FSx file systems as persistent volumes in EKS.

FSx for Lustre

Lustre is a high-performance parallel file system ideal for HPC, ML training, and media processing. It can integrate directly with S3 for data hydration.

Installation

helm repo add aws-fsx-csi-driver https://kubernetes-sigs.github.io/aws-fsx-csi-driver
helm install aws-fsx-csi-driver aws-fsx-csi-driver/aws-fsx-csi-driver \
  --namespace kube-system

StorageClass

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: fsx-lustre
provisioner: fsx.csi.aws.com
parameters:
  subnetId: subnet-0123456789abcdef0
  securityGroupIds: sg-0123456789abcdef0
  deploymentType: PERSISTENT_2
  perUnitStorageThroughput: "200"       # MB/s per TiB
  dataCompressionType: LZ4
volumeBindingMode: WaitForFirstConsumer

S3 Integration

FSx for Lustre can automatically import data from S3 and export results back:

parameters:
  subnetId: subnet-0123456789abcdef0
  securityGroupIds: sg-0123456789abcdef0
  s3ImportPath: s3://my-training-data
  s3ExportPath: s3://my-training-results
  autoImportPolicy: NEW_CHANGED_DELETED
  deploymentType: SCRATCH_2

Use Case: ML Training

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: training-data
spec:
  accessModes:
    - ReadWriteMany
  storageClassName: fsx-lustre
  resources:
    requests:
      storage: 1200Gi    # FSx Lustre minimum: 1.2 TiB

FSx for NetApp ONTAP

Enterprise NAS with NFS/SMB support — useful for workloads migrating from on-premises NetApp:

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: fsx-ontap
provisioner: fsx.csi.aws.com
parameters:
  fileSystemId: fs-0123456789abcdef0
  volumeType: ONTAP

ONTAP Features

Feature Description
Multi-protocol NFS and SMB simultaneously
Snapshots NetApp Snapshot (space-efficient)
Cloning FlexClone (instant, zero-copy)
Tiering Automatic data tiering to S3
Deduplication Inline deduplication and compression

When to Use Which Storage

Storage Best For Access Mode Min Size
EBS (gp3) Databases, single-pod apps RWO 1 GiB
EFS Shared files, CMS, Fargate RWX Elastic
FSx Lustre HPC, ML training, media RWX 1.2 TiB
FSx ONTAP Enterprise NAS migration RWO/RWX 1 TiB
FSx OpenZFS Low-latency shared storage RWX 64 GiB

Best Practices

  1. Use FSx Lustre for ML training — direct S3 integration eliminates data copy
  2. Choose SCRATCH_2 for ephemeral workloads — cheaper, no data replication
  3. Choose PERSISTENT_2 for long-running storage — data replicated within AZ
  4. Consider data compression — LZ4 reduces storage costs with minimal CPU overhead
  5. Right-size your file system — FSx pricing is based on provisioned capacity, not used capacity

Updated:

You may also enjoy