awsdocs / amazon-efs-user-guide

The open source version of the Amazon Elastic File System (EFS) User Guide. You can submit feedback & requests for changes by submitting issues in this repo or by making proposed changes & submitting a pull request.

NOTICE

This repository is archived, read-only, and no longer updated. For more information, read the announcement on the AWS News Blog.

You can find up-to-date AWS technical documentation on the AWS Documentation website, where you can also submit feedback and suggestions for improvement.

amazon-efs-user-guide's Issues

Please include a section on how to use EFS cross-region

In the below Documentation link, please include an additional section on how to mount an efs from completely different region.

This Issue is also related to one of the closed Issue
where changing the region under "/etc/amazon/efs/efs-utils.conf" solved the issue.

Joe from Amazon EFS team, did acknowledge that documentation may be unclear, but even after 1 year this section of this document is still confusing. I had to spent hours figure out that why it was not working. Hope this would be addressed at the earliest.

RHEL6.9 version mismatch

The latest kernel version available is 2.6.32-696.30.1.el6.x86_64 (see https://access.redhat.com/articles/3078#RHEL6)
While the document mentions "kernel 2.6.32-704 or newer"

File Sync API

I'm not sure if this is the best place to ask this question but I can't seem to find information anywhere else. Is there a way to do something like this but using File Sync? We need to automate the sync without using the console.

For example, I want to have a Lambda kick off and create a sync task (and ideally start the sync agent first so we don't have to run it running constantly). It seems there isn't a CLI or JS SDK method to accomplish this.

Docs for mounting EFS via Access Points - with cloudformation

Is there any documentation available on how to mount an EFS via an access point using cloudformation code?

Pls document inode limit

We're currently a GCP NFS user (filestore version). They call out increasing FS size to also increase inode count

https://cloud.google.com/filestore/docs/editing-instances

Can you pls call out the same for EFS. We're creating an AWS support ticket for now to get this info, but it would help for many usecases to have this explicitly documented (we run into this limit in our usecase - lots of tiny files).

Mixed units of metric and IEC

On https://docs.aws.amazon.com/efs/latest/ug/performance.html units of IEC (1024^x) are used. But in several paragraphs the units used are the old style. For example:

When burst credits are available, a file system can drive up to 100 MBps per terabyte (TB) of storage, with a minimum of 100 MBps. If no burst credits are available, a file system can drive up to 50 MBps per TB of storage with a minimum of 1 MBps.

Since this document makes use of mebibytes in other parts one could assume that above sentence refers to the metric version of MBps and not the legacy meaning with 1024^2 as a factor.

I suggest to consistently use the IEC units.

Testing

Testing

EFS FQDN fails to resolve when storageClass is missing (undocumented scenario)

For this section of documentation : https://github.com/awsdocs/amazon-efs-user-guide/blob/main/doc_source/troubleshooting-efs-mounting.md#file-system-mount-using-dns-name-fails

This can fail when you are using EFS encryption-in-transit and do NOT have the storageClass ( https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/examples/kubernetes/encryption_in_transit/specs/storageclass.yaml ) in the cluster, and the error can seem nonsense. The error comes from the pod trying to use the PV/PVC, saying mount failed due to unable to resolve the FQDN and then mentions a seemingly irrelevant ARN for a role that doesn't have the permission for resolution, but isn't related to EFS or the EFS.CSI.Driver at all.

In my scenario, when I added this storageClass, the EFS interfacing immediately worked.

This is an undocumented solution so far as I can tell, and this troubleshooting documentation should add this along the lines of "If you're using EFS encryption-in-transit, check that your storageClass is correctly installed" and link to the correct one.

Honestly though, my confusion also was I wasn't sure if the storageClass was required, since the examples without EFS encryption-in-transit have blank storageClass declarations. IMO the storageClass being unclear as to whether you need or don't need it is what tripped me up here, and then a lack of documentation on this facet/scenario kept me stuck.

VPC Peering Clarification

The documentation mentions

A VPC peering connection is a networking connection between two VPCs that enables you to route traffic between them using private Internet Protocol version 4 (IPv4) or Internet Protocol version 6 (IPv6) addresses. For more information on VPC peering, see What is VPC Peering? in the Amazon VPC Peering Guide.

For Amazon EFS, you can work with VPC peering within a single AWS Region when using C5 or M5 instances. However, other VPC private connectivity mechanisms such as a VPN connection, interregion VPC peering, and intraregion VPC peering using other instance types are not supported.

There is no mention of how to actually get this setup working. An attempt at this ends with the DNS of the filesystem failing to resolve in the peered VPC.

The documentation should explain EFS VPC Peering better.