CSI Proxy (which might be more aptly named "csi-node-proxy") is a binary that exposes a set of gRPC APIs (over named pipes) around local storage operations for nodes in Windows. A container, such as CSI node plugins, can mount the named pipes depending on operations it wants to exercise on the host and invoke the APIs. This allows a storage plugin to run as if were a CSI plugin on linux which have the ability to perform "privileged" actions on the windows host OS.
In the above diagram, there are 3 communication channels being utilized for CSI proxy:
csi.sock: This usually lives at c:/var/lib/kubelet/plugins/csi-plugin/csi.sock, and is used for the host OS to communicate with the CSI node plugincsi-plugin-reg.sock: This lives here, c:/var/lib/kubelet/plugins_registry/csi-plugin-reg.sock, and is used for the kubelet itself to get metadata as needed from the csi plugin .\\.\pipe\csi-proxy: This will live in a pipe, i.e. at \.\pipe/csi-proxy-api-group-v-... : upon receipt of specific requests from the Node plugin, it carries out actions on the node plugins behalf. Since it doesn't execute arbitrary commands on the host, its more secure then simply being a black-box proxy to run host commands.
Each named pipe will support a specific version of an API (e.g. v1alpha1, v1beta1, v1beta2, v1) that targets a specific area of storage (e.g. disk, volume, file, SMB, iSCSI).
CSI drivers are recommended to be deployed as containers. Node plugin containers need to run with privileges to perform storage related operations. However, Windows does not support privileged containers currently. With CSIProxy, the node plugins can now be deployed as unprivileged pods that use the proxy to perform privileged storage operations on the node. Kubernetes administrators will need to install and maintain csi-proxy.exe on all Windows nodes in a manner similar to kubelet.exe.
Recommended K8s Version: 1.18
CSI Proxy is in a stable status (GA Blogpost), the latest versions of the API Groups are:
| API Group | Latest Version | API Docs |
|---|---|---|
| Disk | v1 | link |
| Filesystem | v1 | link |
| SMB | v1 | link |
| Volume | v1 | link |
| iSCSI | v1alpha2 | link to proto |
| System | v1alpha1 | link to proto |
csi-proxy.exe binary has not published yet. You need to clone the repo and build it. The easiest way to build this binary is using the makefile in this repo.
git clone https://github.com/kubernetes-csi/csi-proxy
sudo make build
On successful execution of make build, the output binary csi-proxy.exe will be available under /bin directory.
csi-proxy.exe can be installed and run as binary or run as a Windows service on each Windows node. See the following as an example to run CSI Proxy as a web service.
$flags = "-windows-service -log_file=C:\etc\kubernetes\logs\csi-proxy.log -logtostderr=false"
sc.exe create csiproxy start= "auto" binPath= "C:\etc\kubernetes\node\bin\csi-proxy.exe $flags"
sc.exe failure csiproxy reset= 0 actions= restart/10000
sc.exe start csiproxy
If you are using kube-up to start a Windows cluster, node startup script will automatically run csi-proxy as a service. For GKE 1.18+, csi-proxy will be installed automatically.
--kubelet-path: This is the prefix path of the kubelet path directory in the host file system (C:\var\lib\kubeletis used by default).--working-dir(repeated flag): Prefix path where CSI Proxy is allowed to make privileged operations in the host file system (no value by default).
Deploy and start csiproxy.exe on all Windows hosts in the cluster. Next, the named pipes can be mounted in a CSI node plugin DaemonSet YAML in the following manner:
kind: DaemonSet
apiVersion: apps/v1
metadata:
name: csi-storage-node-win
spec:
selector:
matchLabels:
app: csi-driver-win
template:
metadata:
labels:
app: csi-driver-win
spec:
serviceAccountName: csi-node-sa
tolerations:
- key: "node.kubernetes.io/os"
operator: "Equal"
value: "win1809"
effect: "NoSchedule"
nodeSelector:
kubernetes.io/os: windows
containers:
- name: csi-driver-registrar
image: k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.1.0
args:
- "--v=5"
- "--csi-address=unix://C:\\csi\\csi.sock"
- "--kubelet-registration-path=C:\\var\\lib\\kubelet\\plugins\\pd.csi.storage.gke.io\\csi.sock"
env:
- name: KUBE_NODE_NAME
valueFrom:
fieldRef:
fieldPath: spec.nodeName
volumeMounts:
- name: plugin-dir
mountPath: C:\csi
- name: registration-dir
mountPath: C:\registration
- name: csi-driver
# placeholder, use your CSI driver
image: org/csi-driver:win-v1
args:
- "--v=5"
- "--endpoint=unix:/csi/csi.sock"
volumeMounts:
- name: kubelet-dir
mountPath: C:\var\lib\kubelet
- name: plugin-dir
mountPath: C:\csi
- name: csi-proxy-disk-pipe
mountPath: \\.\pipe\csi-proxy-disk-v1
- name: csi-proxy-volume-pipe
mountPath: \\.\pipe\csi-proxy-volume-v1
- name: csi-proxy-filesystem-pipe
mountPath: \\.\pipe\csi-proxy-filesystem-v1
- name: csi-proxy-smb-pipe
mountPath: \\.\pipe\csi-proxy-smb-v1
volumes:
- name: csi-proxy-disk-pipe
hostPath:
path: \\.\pipe\csi-proxy-disk-v1
type: ""
- name: csi-proxy-volume-pipe
hostPath:
path: \\.\pipe\csi-proxy-volume-v1
type: ""
- name: csi-proxy-filesystem-pipe
hostPath:
path: \\.\pipe\csi-proxy-filesystem-v1
type: ""
- name: csi-proxy-smb-pipe
hostPath:
path: \\.\pipe\csi-proxy-smb-v1
type: ""
- name: registration-dir
hostPath:
path: C:\var\lib\kubelet\plugins_registry\
type: Directory
- name: kubelet-dir
hostPath:
path: C:\var\lib\kubelet\
type: Directory
- name: plugin-dir
hostPath:
path: C:\var\lib\kubelet\plugins\csi.org.io\
type: DirectoryOrCreate
Check out development.md for instructions to set up a development environment to run CSI Proxy.
Learn how to engage with the Kubernetes community on the community page.
You can reach the maintainers of this project at:
-
SMB CSI Driver. To see specifically how this driver is invoked, you can look at https://github.com/kubernetes-csi/csi-driver-smb/blob/master/pkg/mounter/safe_mounter_windows.go.
-
Azure File CSI Driver. See https://github.com/kubernetes-sigs/azurefile-csi-driver/blob/master/pkg/mounter/safe_mounter_windows.go as an example of the invocation path
Participation in the Kubernetes community is governed by the Kubernetes Code of Conduct.