Cache and print devices for debugging future outages

cmd/gce-pd-csi-driver/main.go

@@ -34,6 +34,7 @@ import (
 	gce "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/gce-cloud-provider/compute"
 	metadataservice "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/gce-cloud-provider/metadata"
 	driver "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/gce-pd-csi-driver"
+	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/linkcache"
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/metrics"
 	mountmanager "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/mount-manager"
 )
@@ -97,6 +98,8 @@ var (
 
 	diskTopology = flag.Bool("disk-topology", false, "If set to true, the driver will add a disk-type.gke.io/[disk-type] topology label when the StorageClass has the use-allowed-disk-topology parameter set to true. That topology label is included in the Topologies returned in CreateVolumeResponse. This flag is disabled by default.")
 
+	diskCacheSyncPeriod = flag.Duration("disk-cache-sync-period", 10*time.Minute, "Period for the disk cache to check the /dev/disk/by-id/ directory and evaluate the symlinks")
+
 	enableDiskSizeValidation = flag.Bool("enable-disk-size-validation", false, "If set to true, the driver will validate that the requested disk size is matches the physical disk size. This flag is disabled by default.")
 
 	version string
@@ -278,6 +281,13 @@ func handle() {
 			klog.Fatalf("Failed to get node info from API server: %v", err.Error())
 		}
 
+		deviceCache, err := linkcache.NewDeviceCacheForNode(ctx, *diskCacheSyncPeriod, *nodeName, driverName, deviceUtils)
+		if err != nil {
+			klog.Warningf("Failed to create device cache: %v", err.Error())
+		} else {
+			go deviceCache.Run(ctx)
+		}
+
 		// TODO(2042): Move more of the constructor args into this struct
 		nsArgs := &driver.NodeServerArgs{
 			EnableDeviceInUseCheck:   *enableDeviceInUseCheck,
@@ -286,6 +296,7 @@ func handle() {
 			DataCacheEnabledNodePool: isDataCacheEnabledNodePool,
 			SysfsPath:                "/sys",
 			MetricsManager:           metricsManager,
+			DeviceCache:              deviceCache,
 		}
 		nodeServer = driver.NewNodeServer(gceDriver, mounter, deviceUtils, meta, statter, nsArgs)
 
@@ -300,9 +311,10 @@ func handle() {
 				if err := setupDataCache(ctx, *nodeName, nodeServer.MetadataService.GetName()); err != nil {
 					klog.Errorf("Data Cache setup failed: %v", err)
 				}
-				go driver.StartWatcher(*nodeName)
+				go driver.StartWatcher(ctx, *nodeName)
 			}
 		}
+
 	}
 
 	err = gceDriver.SetupGCEDriver(driverName, version, extraVolumeLabels, extraTags, identityServer, controllerServer, nodeServer)

deploy/kubernetes/base/controller/controller.yaml

@@ -145,6 +145,7 @@ spec:
             - "--supports-dynamic-iops-provisioning=hyperdisk-balanced,hyperdisk-extreme"
             - "--supports-dynamic-throughput-provisioning=hyperdisk-balanced,hyperdisk-throughput,hyperdisk-ml"
             - --enable-data-cache
+            - --run-node-service=false
             - --enable-multitenancy
           command:
             - /gce-pd-csi-driver

deploy/kubernetes/images/stable-master/image.yaml

@@ -46,6 +46,6 @@ imageTag:
   name: gke.gcr.io/gcp-compute-persistent-disk-csi-driver
   # pdImagePlaceholder in test/k8s-integration/main.go is updated automatically with the newTag
   newName: registry.k8s.io/cloud-provider-gcp/gcp-compute-persistent-disk-csi-driver
-  newTag: "v1.17.3"
+  newTag: "v1.17.4"
 ---
 

manifest_osversion.sh

@@ -4,7 +4,7 @@ set -o xtrace
 
 # The following is a workaround for issue https://github.com/moby/moby/issues/41417
 # to manually inserver os.version information into docker manifest file
-# TODO: once docker manifest annotation for os.versions is availabler for the installed docker here,
+# TODO: once docker manifest annotation for os.versions is available for the installed docker here,
 # replace the following with annotation approach. https://github.com/docker/cli/pull/2578
 
 export DOCKER_CLI_EXPERIMENTAL=enabled
@@ -14,7 +14,7 @@ IFS=', ' read -r -a imagetags <<< "$WINDOWS_IMAGE_TAGS"
 IFS=', ' read -r -a baseimages <<< "$WINDOWS_BASE_IMAGES"
 MANIFEST_TAG=${STAGINGIMAGE}:${STAGINGVERSION}
 
-# translate from image tag to docker manifest foler format
+# translate from image tag to docker manifest folder format
 # e.g., gcr.io_k8s-staging-csi_gce-pd-windows-v2
 manifest_folder=$(echo "${MANIFEST_TAG}" | sed "s|/|_|g" | sed "s/:/-/")
 echo ${manifest_folder}

pkg/gce-pd-csi-driver/cache.go

@@ -7,19 +7,14 @@ import (
 	"regexp"
 	"strconv"
 	"strings"
-	"time"
 
 	csi "github.com/container-storage-interface/spec/lib/go/csi"
 	fsnotify "github.com/fsnotify/fsnotify"
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/status"
-	v1 "k8s.io/api/core/v1"
-	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	"k8s.io/apimachinery/pkg/util/wait"
-	"k8s.io/client-go/kubernetes"
-	"k8s.io/client-go/rest"
 	"k8s.io/klog/v2"
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/common"
+	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/k8sclient"
 )
 
 const (
@@ -257,18 +252,11 @@ func ValidateDataCacheConfig(dataCacheMode string, dataCacheSize string, ctx con
 }
 
 func GetDataCacheCountFromNodeLabel(ctx context.Context, nodeName string) (int, error) {
-	cfg, err := rest.InClusterConfig()
-	if err != nil {
-		return 0, err
-	}
-	kubeClient, err := kubernetes.NewForConfig(cfg)
-	if err != nil {
-		return 0, err
-	}
-	node, err := getNodeWithRetry(ctx, kubeClient, nodeName)
+	node, err := k8sclient.GetNodeWithRetry(ctx, nodeName)
 	if err != nil {
 		return 0, err
 	}
+
 	if val, found := node.GetLabels()[fmt.Sprintf(common.NodeLabelPrefix, common.DataCacheLssdCountLabel)]; found {
 		dataCacheCount, err := strconv.Atoi(val)
 		if err != nil {
@@ -280,30 +268,6 @@ func GetDataCacheCountFromNodeLabel(ctx context.Context, nodeName string) (int,
 	return 0, nil
 }
 
-func getNodeWithRetry(ctx context.Context, kubeClient *kubernetes.Clientset, nodeName string) (*v1.Node, error) {
-	var nodeObj *v1.Node
-	backoff := wait.Backoff{
-		Duration: 1 * time.Second,
-		Factor:   2.0,
-		Steps:    5,
-	}
-	err := wait.ExponentialBackoffWithContext(ctx, backoff, func(_ context.Context) (bool, error) {
-		node, err := kubeClient.CoreV1().Nodes().Get(ctx, nodeName, metav1.GetOptions{})
-		if err != nil {
-			klog.Warningf("Error getting node %s: %v, retrying...\n", nodeName, err)
-			return false, nil
-		}
-		nodeObj = node
-		klog.V(4).Infof("Successfully retrieved node info %s\n", nodeName)
-		return true, nil
-	})
-
-	if err != nil {
-		klog.Errorf("Failed to get node %s after retries: %v\n", nodeName, err)
-	}
-	return nodeObj, err
-}
-
 func FetchRaidedLssdCountForDatacache() (int, error) {
 	raidedPath, err := fetchRAIDedLocalSsdPath()
 	if err != nil {
@@ -617,7 +581,7 @@ func InitializeDataCacheNode(nodeId string) error {
 	return nil
 }
 
-func StartWatcher(nodeName string) {
+func StartWatcher(ctx context.Context, nodeName string) {
 	dirToWatch := "/dev/"
 	watcher, err := fsnotify.NewWatcher()
 	if err != nil {
@@ -632,17 +596,20 @@ func StartWatcher(nodeName string) {
 	}
 	errorCh := make(chan error, 1)
 	// Handle the error received from the watcher goroutine
-	go watchDiskDetaches(watcher, nodeName, errorCh)
+	go watchDiskDetaches(ctx, watcher, nodeName, errorCh)
 
 	select {
 	case err := <-errorCh:
 		klog.Errorf("watcher encountered an error: %v", err)
 	}
 }
 
-func watchDiskDetaches(watcher *fsnotify.Watcher, nodeName string, errorCh chan error) error {
+func watchDiskDetaches(ctx context.Context, watcher *fsnotify.Watcher, nodeName string, errorCh chan error) error {
 	for {
 		select {
+		case <-ctx.Done():
+			klog.Infof("Context done, stopping watcher")
+			return nil
 		// watch for errors
 		case err := <-watcher.Errors:
 			errorCh <- fmt.Errorf("disk update event errored: %v", err)

pkg/gce-pd-csi-driver/gce-pd-driver.go

@@ -159,6 +159,7 @@ func NewNodeServer(gceDriver *GCEDriver, mounter *mount.SafeFormatAndMount, devi
 		DataCacheEnabledNodePool: args.DataCacheEnabledNodePool,
 		SysfsPath:                args.SysfsPath,
 		metricsManager:           args.MetricsManager,
+		DeviceCache:              args.DeviceCache,
 	}
 }
 
@@ -185,7 +186,7 @@ func (gceDriver *GCEDriver) Run(endpoint string, grpcLogCharCap int, enableOtelT
 	maxLogChar = grpcLogCharCap
 
 	klog.V(4).Infof("Driver: %v", gceDriver.name)
-	//Start the nonblocking GRPC
+	// Start the nonblocking GRPC
 	s := NewNonBlockingGRPCServer(enableOtelTracing, metricsManager)
 	// TODO(#34): Only start specific servers based on a flag.
 	// In the future have this only run specific combinations of servers depending on which version this is.

pkg/gce-pd-csi-driver/node.go

@@ -32,14 +32,14 @@ import (
 
 	csi "github.com/container-storage-interface/spec/lib/go/csi"
 
-	"k8s.io/client-go/kubernetes"
-	"k8s.io/client-go/rest"
 	"k8s.io/klog/v2"
 	"k8s.io/mount-utils"
 
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/common"
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/deviceutils"
 	metadataservice "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/gce-cloud-provider/metadata"
+	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/k8sclient"
+	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/linkcache"
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/metrics"
 	mountmanager "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/mount-manager"
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/resizefs"
@@ -80,6 +80,8 @@ type GCENodeServer struct {
 	csi.UnimplementedNodeServer
 
 	metricsManager *metrics.MetricsManager
+	// A cache of the device paths for the volumes that are attached to the node.
+	DeviceCache *linkcache.DeviceCache
 }
 
 type NodeServerArgs struct {
@@ -97,6 +99,7 @@ type NodeServerArgs struct {
 	SysfsPath string
 
 	MetricsManager *metrics.MetricsManager
+	DeviceCache    *linkcache.DeviceCache
 }
 
 var _ csi.NodeServer = &GCENodeServer{}
@@ -545,6 +548,13 @@ func (ns *GCENodeServer) NodeStageVolume(ctx context.Context, req *csi.NodeStage
 		}
 	}
 
+	if ns.DeviceCache != nil {
+		err = ns.DeviceCache.AddVolume(volumeID)
+		if err != nil {
+			klog.Warningf("Error adding volume %s to cache: %v", volumeID, err)
+		}
+	}
+
 	klog.V(4).Infof("NodeStageVolume succeeded on %v to %s", volumeID, stagingTargetPath)
 	return &csi.NodeStageVolumeResponse{}, nil
 }
@@ -663,6 +673,11 @@ func (ns *GCENodeServer) NodeUnstageVolume(ctx context.Context, req *csi.NodeUns
 			return nil, status.Errorf(codes.DataLoss, "Failed to cleanup cache for volume %s: %v", volumeID, err)
 		}
 	}
+
+	if ns.DeviceCache != nil {
+		ns.DeviceCache.RemoveVolume(volumeID)
+	}
+
 	klog.V(4).Infof("NodeUnstageVolume succeeded on %v from %s", volumeID, stagingTargetPath)
 	return &csi.NodeUnstageVolumeResponse{}, nil
 }
@@ -940,15 +955,7 @@ func (ns *GCENodeServer) GetVolumeLimits(ctx context.Context) (int64, error) {
 }
 
 func GetAttachLimitsOverrideFromNodeLabel(ctx context.Context, nodeName string) (int64, error) {
-	cfg, err := rest.InClusterConfig()
-	if err != nil {
-		return 0, err
-	}
-	kubeClient, err := kubernetes.NewForConfig(cfg)
-	if err != nil {
-		return 0, err
-	}
-	node, err := getNodeWithRetry(ctx, kubeClient, nodeName)
+	node, err := k8sclient.GetNodeWithRetry(ctx, nodeName)
 	if err != nil {
 		return 0, err
 	}

pkg/gce-pd-csi-driver/node_test.go

@@ -36,6 +36,7 @@ import (
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/common"
 	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/deviceutils"
 	metadataservice "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/gce-cloud-provider/metadata"
+	"sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/linkcache"
 	mountmanager "sigs.k8s.io/gcp-compute-persistent-disk-csi-driver/pkg/mount-manager"
 )
 
@@ -46,11 +47,13 @@ const (
 )
 
 func getTestGCEDriver(t *testing.T) *GCEDriver {
-	return getCustomTestGCEDriver(t, mountmanager.NewFakeSafeMounter(), deviceutils.NewFakeDeviceUtils(false), metadataservice.NewFakeService(), &NodeServerArgs{})
+	return getCustomTestGCEDriver(t, mountmanager.NewFakeSafeMounter(), deviceutils.NewFakeDeviceUtils(false), metadataservice.NewFakeService(), &NodeServerArgs{
+		DeviceCache: linkcache.NewTestDeviceCache(1*time.Minute, linkcache.NewTestNodeWithVolumes([]string{defaultVolumeID})),
+	})
 }
 
-func getTestGCEDriverWithCustomMounter(t *testing.T, mounter *mount.SafeFormatAndMount) *GCEDriver {
-	return getCustomTestGCEDriver(t, mounter, deviceutils.NewFakeDeviceUtils(false), metadataservice.NewFakeService(), &NodeServerArgs{})
+func getTestGCEDriverWithCustomMounter(t *testing.T, mounter *mount.SafeFormatAndMount, args *NodeServerArgs) *GCEDriver {
+	return getCustomTestGCEDriver(t, mounter, deviceutils.NewFakeDeviceUtils(false), metadataservice.NewFakeService(), args)
 }
 
 func getCustomTestGCEDriver(t *testing.T, mounter *mount.SafeFormatAndMount, deviceUtils deviceutils.DeviceUtils, metaService metadataservice.MetadataService, args *NodeServerArgs) *GCEDriver {
@@ -190,7 +193,9 @@ func TestNodeGetVolumeStats(t *testing.T) {
 			}
 
 			mounter := mountmanager.NewFakeSafeMounterWithCustomExec(&testingexec.FakeExec{CommandScript: actionList})
-			gceDriver := getTestGCEDriverWithCustomMounter(t, mounter)
+			gceDriver := getTestGCEDriverWithCustomMounter(t, mounter, &NodeServerArgs{
+				DeviceCache: linkcache.NewTestDeviceCache(1*time.Minute, linkcache.NewTestNodeWithVolumes([]string{tc.volumeID})),
+			})
 			ns := gceDriver.ns
 
 			req := &csi.NodeGetVolumeStatsRequest{
@@ -1300,7 +1305,9 @@ func TestNodeStageVolume(t *testing.T) {
 				))
 			}
 			mounter := mountmanager.NewFakeSafeMounterWithCustomExec(&testingexec.FakeExec{CommandScript: actionList, ExactOrder: true})
-			gceDriver := getTestGCEDriverWithCustomMounter(t, mounter)
+			gceDriver := getTestGCEDriverWithCustomMounter(t, mounter, &NodeServerArgs{
+				DeviceCache: linkcache.NewTestDeviceCache(1*time.Minute, linkcache.NewTestNodeWithVolumes([]string{volumeID})),
+			})
 			ns := gceDriver.ns
 			ns.SysfsPath = tempDir + "/sys"
 			_, err := ns.NodeStageVolume(context.Background(), tc.req)

pkg/k8sclient/node.go

@@ -0,0 +1,49 @@
+package k8sclient
+
+import (
+	"context"
+	"time"
+
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/wait"
+	"k8s.io/client-go/kubernetes"
+	"k8s.io/client-go/rest"
+	"k8s.io/klog/v2"
+)
+
+func GetNodeWithRetry(ctx context.Context, nodeName string) (*v1.Node, error) {
+	cfg, err := rest.InClusterConfig()
+	if err != nil {
+		return nil, err
+	}
+	kubeClient, err := kubernetes.NewForConfig(cfg)
+	if err != nil {
+		return nil, err
+	}
+	return getNodeWithRetry(ctx, kubeClient, nodeName)
+}
+
+func getNodeWithRetry(ctx context.Context, kubeClient *kubernetes.Clientset, nodeName string) (*v1.Node, error) {
+	var nodeObj *v1.Node
+	backoff := wait.Backoff{
+		Duration: 1 * time.Second,
+		Factor:   2.0,
+		Steps:    5,
+	}
+	err := wait.ExponentialBackoffWithContext(ctx, backoff, func(_ context.Context) (bool, error) {
+		node, err := kubeClient.CoreV1().Nodes().Get(ctx, nodeName, metav1.GetOptions{})
+		if err != nil {
+			klog.Warningf("Error getting node %s: %v, retrying...\n", nodeName, err)
+			return false, nil
+		}
+		nodeObj = node
+		klog.V(4).Infof("Successfully retrieved node info %s\n", nodeName)
+		return true, nil
+	})
+
+	if err != nil {
+		klog.Errorf("Failed to get node %s after retries: %v\n", nodeName, err)
+	}
+	return nodeObj, err
+}