Driver enforces a maximum AccessPointPerFsLimit of 1000 despite the default (and customisable) quota being 10000

[pmjacinto]

/kind bug

What happened?

EFS default quota for access points per FS was bumped from 1000 to 10000 last year: https://aws.amazon.com/about-aws/whats-new/2025/02/amazon-efs-access-points-file-system/

Note that this quota is customisable upon request so it can be larger: https://docs.aws.amazon.com/efs/latest/ug/limits.html#soft-limits

When using dynamic allocation with gid assignment, the code only allows for a pool of 1000 IDs effectively disallowing more than 1000 APs per FS. There are also multiple EFS API calls which assume a max of 1000 APs.

https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/pkg/cloud/cloud.go#L44
https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/pkg/driver/gid_allocator.go#L70-L74
https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/pkg/cloud/cloud.go#L281
https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/pkg/cloud/cloud.go#L317

What you expected to happen?

Ideally, AccessPointPerFsLimit should be customisable to reflect it being a soft quota. Or, at a minimum, the value should be updated from 1000 to 10000 to reflect the current default quota.

How to reproduce it (as minimally and precisely as possible)?

Create more than 1000 PVCs for the same EFS FS using dynamic provisioning and dynamic gid allocation.

Anything else we need to know?:

Environment

• Kubernetes version (use kubectl version): 1.33
• Driver version: latest