Hey @MarkSpencerTan ,

Would a provisioning mode which would provision file systems in place of access points help your use case? #310

@kbasv we would like to keep the file system count at minimum so we can have a price friendly backup plan/long term option

@kbasv also if the EFS CSI driver can support reusing just one access point and then creating subdirectories inside of it per pvc, that would be another way we can just use one EFS... not sure if that's possible already with the current implementation since I have seen mentions of mountOptions and volumeHandle examples where you can specify subdirectories inside an access point.

@kbasv we need this ability too and we would love to get this escalated, we have over 120 mounts

I checked with AWS, it doesn't seem like they support service limit increase for this.

I can contrib code if you show me where it needs a fix

The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle stale
• Mark this issue or PR as rotten with /lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Reopen this issue or PR with /reopen
• Mark this issue or PR as fresh with /remove-lifecycle rotten
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/close

@k8s-triage-robot: Closing this issue.

I know this issue is closed, but I will like to make a comment for future Googlers ;) Creating an Access Points for each PVC is in my opinion a huge shortcoming for the dynamic provisioner. In one of the K8s clusters I manage we have more then 1000 ReadWriteMany PVCs (and more then 4500 PODs). The efs csi-driver dynamic provisioner doesn't work for us due to the Access Points limitations pr filesystem. Sure, creating more filesystem is possible, but then you also have to create several storage classes and make sure to not have more then 120 PVCs pr Storage Class, not very user friendly. Also the ability to share IOPS and provisioned throughput of one filesystem across all pods save us money. Not all pods do "spike" at the same time.

So in our use case a better model would be that each PVC just get's it's own subfolder (yes I know that an Access Point provide more features when it comes to file owners and access constraints, then a plain subfolder does). There is actually an old, and unfortunately not maintained, provisioner out there that does exactly this. Provision a subfolder in a shared EFS pr PVC, https://github.com/goto-opensource/aws-efs-csi-pv-provisioner

We have been running with this provisioner in a large scale cluster for years, and it's working great. Most of the "hard work" (mount / dismount EFS etc) is anyway handled by the official EFS CSI driver.

@paalkr Your post pretty much exactly matches our experience with the old provisioner and the trouble we are having with the AP limitation since we had to make the switch. A mode that just creates subdirectories would be what we need.

@paalkr @jgoeres it seems like @jonathanrainer is proposing a directory-based approach (not using access points) in PR #732, you may want to check it out!

But in the mean time, the official NFS CSI Driver already uses a directory-based approach (so has no 120 limit), and works with any arbitrary NFS servers (including EFS).