Implement Windows NodePublish/Unpublish

[wongma7]

Is this a bug fix or adding new feature? /feature

What is this PR about? / Why do we need it? In-tree EBS driver supports windows, so too must CSI in case an in-tree EBS volume gets migrated to CSI

I will add build/push job to CI in a separate PR. Since windows images are big, don't want to prematurely upload stuff to registry.

NOTE: this requires a not-yet-released version of csi-proxy with changes from

• Return serial number in ListDiskIDs kubernetes-csi/csi-proxy#116
• Close Open disk file handles kubernetes-csi/csi-proxy#126
• Fix IsVolumeFormatted issue on checking volume type "Unknown" kubernetes-csi/csi-proxy#127
  I will detail this in README so that my test results are reproducible.

TODO:

• basic/minimal e2e test with migration off
• doc/example

What testing is done?
Windows mount/unmount works, data persists.

I will let CI test that my refactoring has not broken Linux.

# build and push image
 docker buildx build --platform windows -f Dockerfile.windows .

Manual test:

# Create EKS cluster with Windows nodes.

# Checkout this PR.

# Install the driver including the windows daemonset
helm upgrade --install aws-ebs-csi-driver --namespace kube-system ./charts/aws-ebs-csi-driver

# Replace the windows daemonset image with the windows image. It's a TODO to make the image multiarch... # public.ecr.aws/b5w6x5z2/aws-ebs-csi-driver:windows
k edit daemonset ebs-csi-node-windows -n kube-system

kubectl apply -f examples/kubernetes/windows/specs/

k exec -it windows-server-iis-7c5fc8f6c5-t5mk9 -- powershell

PS C:\> New-Item -Path data -Name "testfile1.txt" -ItemType "file" -Value "This 
is a text string."


    Directory: C:\data


Mode                LastWriteTime         Length Name
----                -------------         ------ ----
-a----         4/7/2021  12:31 AM             22 testfile1.txt

k delete po windows-server-iis-7c5fc8f6c5-t5mk9

k exec -it windows-server-iis-7c5fc8f6c5-j44qv -- powershell

PS C:\> ls data 


    Directory: C:\data 


Mode                LastWriteTime         Length Name
----                -------------         ------ ----
-a----         4/7/2021  12:31 AM             22 testfile1.txt

Basic/minimal e2e test:

$ ginkgo -nodes=1 -v --focus="External.Storage.*default.fs.*should.store.data" ./tests/e2e-kubernetes/ -- -kubeconfig=$KUBECONFIG -gce-zone=us-west-2a -node-os-distro=windows
...

• [SLOW TEST:144.768 seconds]
External Storage [Driver: ebs.csi.aws.com]
/home/ANT.AMAZON.COM/mattwon/go/src/github.com/kubernetes-sigs/aws-ebs-csi-driver/vendor/k8s.io/kubernetes/test/e2e/storage/external/external.go:169
  [Testpattern: Dynamic PV (default fs)] volumes
  /home/ANT.AMAZON.COM/mattwon/go/src/github.com/kubernetes-sigs/aws-ebs-csi-driver/vendor/k8s.io/kubernetes/test/e2e/storage/testsuites/base.go:129
    should store data
    /home/ANT.AMAZON.COM/mattwon/go/src/github.com/kubernetes-sigs/aws-ebs-csi-driver/vendor/k8s.io/kubernetes/test/e2e/storage/testsuites/volumes.go:151
------------------------------
...
Ran 1 of 481 Specs in 144.773 seconds
SUCCESS! -- 1 Passed | 0 Failed | 0 Pending | 480 Skipped
PASS

[coveralls]

Coverage increased (+0.06%) to 79.312% when pulling 1631174 on wongma7:windowsforreal into 95ab71a on kubernetes-sigs:master.

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: wongma7

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [wongma7]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[wongma7]

turns out to be issue in csi-proxy, fix here: kubernetes-csi/csi-proxy#126

[wongma7]

@AndyXiangLi PTAL

We have to settle for my manual testing because there is no EKS windows AMI with csi-proxy.exe (with the necessary patches) installed meaning it's IMO too difficult to write an automated test. (somethingn eeds to apply the patches, build the binary, install the service into the node ami, all as prerequisite for the test, as described in my example README) However we will need to have CI exercising automatic tests before kubernetes 1.23-ish? to make sure the GA driver + windows + GA migration all works together.

[AndyXiangLi]

I agree that we should have CI ready to cover windows use case by 1.23.
To set the expectation right, should we update the readme and mention resizing is not supported on windows?

[wongma7]

I added one more sentence to the windows example README: "Only basic read/write (mount/unmount and attach/detach) functionality has been tested, other features like resize don't work yet."

I don't want to mention windows support in top-level README yet because it's not possible without applying patches to csi-proxy.exe, as mentioned in pre-requisites section of windows example README

[wongma7]

/hold

[wongma7]

/hold cancel

[wongma7]

kubernetes-csi/csi-proxy#144

/hold

need to retest again after kubernetes-csi/csi-proxy#144 gets fixed

[wongma7]

/hold cancel

all the required changes to csi-proxy have merged and I tested the PR with them

[wongma7]

/test pull-aws-ebs-csi-driver-external-test

[nckturner]

What is the difference in terms of function between what you have and the common util?

[wongma7]

it's copied, but IIRC it existed in kubernetes/kubernetes not kubernetes/mount-util at the time so xiangli and I agreed to copy and leave as TODO , instead of depending on kubernetes/kubernetes which is generally bad idea, see

"
TODO: Upstream is planning on move the common resize code to kubernetes/mount-util, #99223, we should use the resize function from there instead of maintaining our own once the changes are available"
#753

[nckturner]

/lgtm