feat: CSI RClone Improvements

.devcontainer/devcontainer.json

@@ -1,9 +1,13 @@
 {
     "name": "CSI rclone devcontainer",
     "image": "mcr.microsoft.com/devcontainers/base:bookworm",
+    "remoteUser": "root",
+    "containerUser": "root",
     "features": {
         "ghcr.io/devcontainers/features/git:1": {},
-        "ghcr.io/devcontainers/features/go:1": {},
+        "ghcr.io/devcontainers/features/go:1": {
+            "version": "latest"
+        },
         "ghcr.io/devcontainers-extra/features/apt-packages:1": {
             "packages": "fuse3"
         },

.devcontainer/rclone/install.sh

@@ -19,3 +19,6 @@ rm -rf /tmp/rclone
 # Fix the $GOPATH folder
 chown -R "${USERNAME}:golang" /go
 chmod -R g+r+w /go
+
+# Make sure the default folders exists
+mkdir -p /run/csi-rclone

Dockerfile

@@ -2,7 +2,7 @@ ARG RCLONE_IMAGE_REPOSITORY="ghcr.io/swissdatasciencecenter/rclone"
 ARG RCLONE_IMAGE_TAG="sha-308067c"
 FROM ${RCLONE_IMAGE_REPOSITORY}:${RCLONE_IMAGE_TAG} AS rclone
 
-FROM golang:1.23.8-bookworm AS build
+FROM golang:1.25.6-bookworm AS build
 COPY go.mod go.sum ./
 RUN --mount=type=cache,target=/go/pkg/mod \
   go mod download
@@ -23,4 +23,4 @@ EOT
 COPY --from=build /csi-rclone /csi-rclone
 COPY --from=rclone --chmod=755 /rclone /usr/bin/
 
-ENTRYPOINT ["/csi-rclone"]
+ENTRYPOINT ["/csi-rclone"]

cmd/csi-rclone-plugin/main.go

@@ -2,33 +2,35 @@ package main
 
 import (
 	"context"
+	"errors"
 	"flag"
 	"fmt"
 	"os"
 	"os/signal"
-	"syscall"
 	"time"
 
+	"github.com/SwissDataScienceCenter/csi-rclone/pkg/common"
 	"github.com/SwissDataScienceCenter/csi-rclone/pkg/metrics"
 	"github.com/SwissDataScienceCenter/csi-rclone/pkg/rclone"
 	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
 	"k8s.io/klog"
-	mountUtils "k8s.io/mount-utils"
 )
 
-var (
-	endpoint  string
-	nodeID    string
-	cacheDir  string
-	cacheSize string
-	meters    []metrics.Observable
-)
+func exitOnError(err error) {
+	// ParseFlags uses errors to return some status information, ignore it here.
+	if err != nil && !errors.Is(err, pflag.ErrHelp) {
+		klog.Error(err.Error())
+		os.Exit(1)
+	}
+}
 
 func init() {
-	flag.Set("logtostderr", "true")
+	exitOnError(flag.Set("logtostderr", "true"))
 }
 
 func main() {
+	var meters []metrics.Observable
 	metricsServerConfig := metrics.ServerConfig{
 		Host:            "localhost",
 		Port:            9090,
@@ -37,123 +39,49 @@ func main() {
 		ShutdownTimeout: 5 * time.Second,
 		Enabled:         false,
 	}
+	nodeServerConfig := rclone.NodeServerConfig{}
+	controllerServerConfig := rclone.ControllerServerConfig{}
 
 	root := &cobra.Command{
 		Use:   "rclone",
 		Short: "CSI based rclone driver",
 	}
+	// Allow flags to be defined in subcommands, they will be reported at the Execute() step, with the help printed
+	// before exiting.
+	root.FParseErrWhitelist.UnknownFlags = true
+
 	metricsServerConfig.CommandLineParameters(root)
 
 	runCmd := &cobra.Command{
 		Use:   "run",
 		Short: "Start the CSI driver.",
 	}
-	root.AddCommand(runCmd)
+	exitOnError(nodeServerConfig.CommandLineParameters(runCmd, &meters))
+	exitOnError(controllerServerConfig.CommandLineParameters(runCmd, &meters))
 
-	runNode := &cobra.Command{
-		Use:   "node",
-		Short: "Start the CSI driver node service - expected to run in a daemonset on every node.",
-		Run: func(cmd *cobra.Command, args []string) {
-			handleNode()
-		},
-	}
-	runNode.PersistentFlags().StringVar(&nodeID, "nodeid", "", "node id")
-	runNode.MarkPersistentFlagRequired("nodeid")
-	runNode.PersistentFlags().StringVar(&endpoint, "endpoint", "", "CSI endpoint")
-	runNode.MarkPersistentFlagRequired("endpoint")
-	runNode.PersistentFlags().StringVar(&cacheDir, "cachedir", "", "cache dir")
-	runNode.PersistentFlags().StringVar(&cacheSize, "cachesize", "", "cache size")
-	runCmd.AddCommand(runNode)
-	runController := &cobra.Command{
-		Use:   "controller",
-		Short: "Start the CSI driver controller.",
-		Run: func(cmd *cobra.Command, args []string) {
-			handleController()
-		},
-	}
-	runController.PersistentFlags().StringVar(&nodeID, "nodeid", "", "node id")
-	runController.MarkPersistentFlagRequired("nodeid")
-	runController.PersistentFlags().StringVar(&endpoint, "endpoint", "", "CSI endpoint")
-	runController.MarkPersistentFlagRequired("endpoint")
-	runCmd.AddCommand(runController)
+	root.AddCommand(runCmd)
 
 	versionCmd := &cobra.Command{
 		Use:   "version",
 		Short: "Prints information about this version of csi rclone plugin",
 		Run: func(cmd *cobra.Command, args []string) {
-			fmt.Printf("csi-rclone plugin Version: %s", rclone.DriverVersion)
+			fmt.Printf("csi-rclone plugin Version: %s\n", rclone.DriverVersion)
 		},
 	}
 	root.AddCommand(versionCmd)
 
-	root.ParseFlags(os.Args[1:])
+	exitOnError(root.ParseFlags(os.Args[1:]))
 
 	if metricsServerConfig.Enabled {
 		// Gracefully exit the metrics background servers
-		ctx, stop := signal.NotifyContext(context.Background(), syscall.SIGTERM, syscall.SIGINT)
+		ctx, stop := signal.NotifyContext(context.Background(), common.InterruptSignals...)
 		defer stop()
 
 		metricsServer := metricsServerConfig.NewServer(ctx, &meters)
 		go metricsServer.ListenAndServe()
 	}
 
-	if err := root.Execute(); err != nil {
-		fmt.Fprintf(os.Stderr, "%s", err.Error())
-		os.Exit(1)
-	}
+	exitOnError(root.Execute())
 
 	os.Exit(0)
 }
-
-func handleNode() {
-	err := unmountOldVols()
-	if err != nil {
-		klog.Warningf("There was an error when trying to unmount old volumes: %v", err)
-	}
-	d := rclone.NewDriver(nodeID, endpoint)
-	ns, err := rclone.NewNodeServer(d.CSIDriver, cacheDir, cacheSize)
-	if err != nil {
-		panic(err)
-	}
-	meters = append(meters, ns.Metrics()...)
-	d.WithNodeServer(ns)
-	err = d.Run()
-	if err != nil {
-		panic(err)
-	}
-}
-
-func handleController() {
-	d := rclone.NewDriver(nodeID, endpoint)
-	cs := rclone.NewControllerServer(d.CSIDriver)
-	meters = append(meters, cs.Metrics()...)
-	d.WithControllerServer(cs)
-	err := d.Run()
-	if err != nil {
-		panic(err)
-	}
-}
-
-// unmountOldVols is used to unmount volumes after a restart on a node
-func unmountOldVols() error {
-	const mountType = "fuse.rclone"
-	const unmountTimeout = time.Second * 5
-	klog.Info("Checking for existing mounts")
-	mounter := mountUtils.Mounter{}
-	mounts, err := mounter.List()
-	if err != nil {
-		return err
-	}
-	for _, mount := range mounts {
-		if mount.Type != mountType {
-			continue
-		}
-		err := mounter.UnmountWithForce(mount.Path, unmountTimeout)
-		if err != nil {
-			klog.Warningf("Failed to unmount %s because of %v.", mount.Path, err)
-			continue
-		}
-		klog.Infof("Sucessfully unmounted %s", mount.Path)
-	}
-	return nil
-}

deploy/csi-rclone/templates/csi-controller-rclone.yaml

@@ -15,6 +15,8 @@ spec:
     metadata:
       labels:
         app: csi-controller-rclone
+      annotations:
+      {{- toYaml .Values.csiControllerRclone.podAnnotations | nindent 8 }}
     spec:
       serviceAccountName: {{ include "chart.fullname" . }}-controller
       containers:
@@ -54,8 +56,8 @@ spec:
         image: {{ .Values.csiControllerRclone.csiProvisioner.image.repository }}:{{ .Values.csiControllerRclone.csiProvisioner.image.tag | default .Chart.AppVersion }}
         imagePullPolicy: {{ .Values.csiControllerRclone.csiProvisioner.imagePullPolicy }}
         volumeMounts:
-          - name: socket-dir
-            mountPath: /csi
+          - mountPath: /csi
+            name: socket-dir
       - name: rclone
         args:
         - run
@@ -85,7 +87,7 @@ spec:
             fieldRef:
               fieldPath: spec.nodeName
         - name: CSI_ENDPOINT
-          value: "unix://plugin/csi.sock"
+          value: "unix://csi/csi.sock"
         - name: KUBERNETES_CLUSTER_DOMAIN
           value: {{ quote .Values.kubernetesClusterDomain }}
         {{- if .Values.csiControllerRclone.rclone.goMemLimit }}
@@ -114,7 +116,7 @@ spec:
           timeoutSeconds: 3
           periodSeconds: 2
         volumeMounts:
-        - mountPath: /plugin
+        - mountPath: /csi
           name: socket-dir
       - name: liveness-probe
         imagePullPolicy: Always

deploy/csi-rclone/templates/csi-nodeplugin-rclone.yaml

@@ -15,14 +15,16 @@ spec:
       labels:
         app: csi-nodeplugin-rclone
       {{- include "chart.selectorLabels" . | nindent 8 }}
+      annotations:
+      {{- toYaml .Values.csiNodepluginRclone.podAnnotations | nindent 8 }}
     spec:
       serviceAccountName: {{ include "chart.fullname" . }}-nodeplugin
       dnsPolicy: ClusterFirstWithHostNet
       containers:
       - name: node-driver-registrar
         args:
         - --v=5
-        - --csi-address=/plugin/csi.sock
+        - --csi-address=/csi/csi.sock
         - --kubelet-registration-path=/var/lib/kubelet/plugins/{{ .Values.storageClassName }}/csi.sock
         env:
         - name: KUBE_NODE_NAME
@@ -45,17 +47,17 @@ spec:
         resources:
           {{- toYaml .Values.csiNodepluginRclone.rclone.resources | nindent 12 }}
         volumeMounts:
-        - mountPath: /plugin
+        - mountPath: /csi
           name: plugin-dir
         - mountPath: /registration
           name: registration-dir
       - name: liveness-probe
         imagePullPolicy: Always
         image: registry.k8s.io/sig-storage/livenessprobe:v2.15.0
         args:
-        - --csi-address=/plugin/csi.sock
+        - --csi-address=/csi/csi.sock
         volumeMounts:
-        - mountPath: /plugin
+        - mountPath: /csi
           name: plugin-dir
       - name: rclone
         args:
@@ -86,7 +88,7 @@ spec:
             fieldRef:
               fieldPath: spec.nodeName
         - name: CSI_ENDPOINT
-          value: "unix://plugin/csi.sock"
+          value: "unix://csi/csi.sock"
         - name: KUBERNETES_CLUSTER_DOMAIN
           value: {{ quote .Values.kubernetesClusterDomain }}
         - name: DRIVER_NAME
@@ -134,8 +136,10 @@ spec:
           timeoutSeconds: 10
           periodSeconds: 30
         volumeMounts:
-        - mountPath: /plugin
+        - mountPath: /csi
           name: plugin-dir
+        - mountPath: /run/csi-rclone
+          name: node-temp-dir
         - mountPath: /var/lib/kubelet/pods
           mountPropagation: Bidirectional
           name: pods-mount-dir
@@ -154,6 +158,11 @@ spec:
         {{ toYaml . | nindent 8 }}
     {{- end }}
       volumes:
+      - hostPath:
+          # NOTE: We mount on /tmp because we want the saved configuration to not survive a whole node restart.
+          path: /tmp/{{.Release.Namespace}}-{{.Release.Name}}-{{.Release.Revision}}
+          type: DirectoryOrCreate
+        name: node-temp-dir
       - hostPath:
           path: {{ .Values.kubeletDir }}/plugins/{{ .Values.storageClassName }}
           type: DirectoryOrCreate
@@ -167,4 +176,4 @@ spec:
           type: DirectoryOrCreate
         name: registration-dir
       - name: cache-dir
-        emptyDir:
+        emptyDir: {}

deploy/csi-rclone/values.yaml

@@ -1,5 +1,8 @@
 storageClassName: csi-rclone
 csiControllerRclone:
+  podAnnotations:
+    prometheus.io/scrape: "true"
+    prometheus.io/port: "9090"
   csiAttacher:
     image:
       repository: registry.k8s.io/sig-storage/csi-attacher
@@ -22,10 +25,10 @@ csiControllerRclone:
       #   memory: 128Mi
       # requests:
       #   cpu: 100m
-      #   memory: 128M
+      #   memory: 128Mi
     # If set, used to set GOMEMLIMIT, it should be strictly lower than
     # limits.memory to prevent OOMkills
-    goMemLimit: # 115Mi
+    goMemLimit: # 115MiB
     # Prometheus metrics
     metrics:
       enabled: true
@@ -38,6 +41,9 @@ csiControllerRclone:
     annotations: {}
 
 csiNodepluginRclone:
+  podAnnotations:
+    prometheus.io/scrape: "true"
+    prometheus.io/port: "9090"
   nodeDriverRegistrar:
     image:
       repository: registry.k8s.io/sig-storage/csi-node-driver-registrar
@@ -68,7 +74,7 @@ csiNodepluginRclone:
       #   memory: 128Mi
     # If set, used to set GOMEMLIMIT, it should be strictly lower than
     # limits.memory to prevent OOMkills
-    goMemLimit: # 115Mi
+    goMemLimit: # 115MiB
     # Prometheus metrics
     metrics:
       enabled: true