Skip to content

Commit

Permalink
Update docs for APs on same EFS
Browse files Browse the repository at this point in the history
Investigation unveiled the bug reported in
[driver issue #167](kubernetes-sigs/aws-efs-csi-driver#167)
whereby you can't use multiple access points backed by the same EFS volume from
within the same pod. This commit updates the README, TESTING, and TODO docs
accordingly.
  • Loading branch information
2uasimojo committed May 20, 2020
1 parent c860d3d commit 66c820d
Show file tree
Hide file tree
Showing 3 changed files with 10 additions and 6 deletions.
10 changes: 8 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -33,8 +33,9 @@ configured appropriately with respect to VPC, availability zones, etc.
Create a separate [access point](https://docs.aws.amazon.com/efs/latest/ug/create-access-point.html) for each
distinct data store you wish to access from your cluster. Be sure to configure ownership and permissions that
will allow read and/or write access by your pod's `uid`/`gid` as desired.
(Note: there is no need to use separate EFS file systems; distinct access points backed by the same EFS file system
will still appear as distinct data stores.)

**Note**: Until [this driver issue](https://github.com/kubernetes-sigs/aws-efs-csi-driver/issues/167) is resolved,
access points must be backed by separate EFS file systems if they are to be used from the same pod.

### Working with `SharedVolume` resources

Expand Down Expand Up @@ -186,6 +187,11 @@ can leave it in an unusable state, even if the operator is able to resurrect the
The only supported way to delete a `PersistentVolumeClaim` (or `PersistentVolume`) associated with a `SharedVolume`
is to delete the `SharedVolume` and let the operator do the rest.

### Separate access points, separate file systems

Until [this driver issue](https://github.com/kubernetes-sigs/aws-efs-csi-driver/issues/167) is resolved,
access points must be backed by separate EFS file systems if they are to be used from the same pod.

## Under the hood

The operator has two controllers. One monitors the resources necessary to run the
Expand Down
4 changes: 2 additions & 2 deletions TESTING.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,10 +24,10 @@
* Validate shared write access from pods in each namespace ✔
* Multiple APs, same FS (should appear as separate data stores) ✔
* Separate pods ✔
* <font color=red>Single pod with multiple volume defs</font>
* Single pod with multiple volume defs -- [broken](https://github.com/kubernetes-sigs/aws-efs-csi-driver/issues/167)
* Separate APs on separate FSes (should appear as separate data stores) ✔
* Separate pods ✔
* <font color=red>Single pod with multiple volume defs</font>
* Single pod with multiple volume defs

## Weird cases

Expand Down
2 changes: 0 additions & 2 deletions TODO.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,8 +12,6 @@ DESCRIPTION:
<empty>
```

* Get a definitive answer on multiple mounts in the same pod

* Set up permissions so the customer can create `SharedVolume` CRs.
* On that note, is there an RBACism that will forbid them from editing a `SharedVolume` once it's created?
That would make some things easier (like allowing us to remove `uneditSharedVolume`).
Expand Down

0 comments on commit 66c820d

Please sign in to comment.