relax volume lifecycle checks by default

The recently introduced "still in use" check revealed a bug in
Kubernetes (kubernetes/kubernetes#101911). While
the check itself is correct, enabling it would cause a lot of test
flakes. Therefore the check has to be disabled until all Kubernetes
versions that we test against are fixed.

cmd/hostpathplugin/main.go

@@ -52,6 +52,7 @@ func main() {
 	flag.Int64Var(&cfg.MaxVolumesPerNode, "maxvolumespernode", 0, "limit of volumes per node")
 	flag.Var(&cfg.Capacity, "capacity", "Simulate storage capacity. The parameter is <kind>=<quantity> where <kind> is the value of a 'kind' storage class parameter and <quantity> is the total amount of bytes for that kind. The flag may be used multiple times to configure different kinds.")
 	flag.BoolVar(&cfg.EnableAttach, "enable-attach", false, "Enables RPC_PUBLISH_UNPUBLISH_VOLUME capability.")
+	flag.BoolVar(&cfg.CheckVolumeLifecycle, "check-volume-lifecycle", false, "Can be used to turn some violations of the volume lifecycle into warnings instead of failing the incorrect gRPC call. Disabled by default because of https://github.com/kubernetes/kubernetes/issues/101911.")
 	flag.Int64Var(&cfg.MaxVolumeSize, "max-volume-size", 1024*1024*1024*1024, "maximum size of volumes in bytes (inclusive)")
 	flag.BoolVar(&cfg.EnableTopology, "enable-topology", true, "Enables PluginCapability_Service_VOLUME_ACCESSIBILITY_CONSTRAINTS capability.")
 	flag.BoolVar(&cfg.EnableVolumeExpansion, "node-expand-required", true, "Enables NodeServiceCapability_RPC_EXPAND_VOLUME capacity.")

go.mod

@@ -15,6 +15,7 @@ require (
 	google.golang.org/genproto v0.0.0-20201209185603-f92720507ed4 // indirect
 	google.golang.org/grpc v1.34.0
 	k8s.io/apimachinery v0.21.0-alpha.0
+	k8s.io/klog/v2 v2.4.0
 	k8s.io/kubernetes v1.20.0
 	k8s.io/mount-utils v0.20.0 // indirect
 	k8s.io/utils v0.0.0-20201110183641-67b214c5f920

pkg/hostpath/controllerserver.go

@@ -33,6 +33,7 @@ import (
 	"google.golang.org/grpc/status"
 
 	"github.com/container-storage-interface/spec/lib/go/csi"
+	"k8s.io/klog/v2"
 	utilexec "k8s.io/utils/exec"
 
 	"github.com/kubernetes-csi/csi-driver-host-path/pkg/state"
@@ -204,8 +205,12 @@ func (hp *hostPath) DeleteVolume(ctx context.Context, req *csi.DeleteVolumeReque
 	}
 
 	if vol.Attached || !vol.Published.Empty() || !vol.Staged.Empty() {
-		return nil, status.Errorf(codes.Internal, "Volume '%s' is still used (attached: %v, staged: %v, published: %v) by '%s' node",
+		msg := fmt.Sprintf("Volume '%s' is still used (attached: %v, staged: %v, published: %v) by '%s' node",
 			vol.VolID, vol.Attached, vol.Staged, vol.Published, vol.NodeID)
+		if hp.config.CheckVolumeLifecycle {
+			return nil, status.Error(codes.Internal, msg)
+		}
+		klog.Warning(msg)
 	}
 
 	if err := hp.deleteVolume(volId); err != nil {
@@ -340,8 +345,12 @@ func (hp *hostPath) ControllerUnpublishVolume(ctx context.Context, req *csi.Cont
 
 	// Check to see if the volume is staged/published on a node
 	if !vol.Published.Empty() || !vol.Staged.Empty() {
-		return nil, status.Errorf(codes.Internal, "Volume '%s' is still used (staged: %v, published: %v) by '%s' node",
+		msg := fmt.Sprintf("Volume '%s' is still used (staged: %v, published: %v) by '%s' node",
 			vol.VolID, vol.Staged, vol.Published, vol.NodeID)
+		if hp.config.CheckVolumeLifecycle {
+			return nil, status.Error(codes.Internal, msg)
+		}
+		klog.Warning(msg)
 	}
 
 	vol.Attached = false

pkg/hostpath/hostpath.go

@@ -74,6 +74,7 @@ type Config struct {
 	EnableAttach          bool
 	EnableTopology        bool
 	EnableVolumeExpansion bool
+	CheckVolumeLifecycle  bool
 }
 
 var (

vendor/modules.txt

@@ -154,6 +154,7 @@ k8s.io/apiserver/pkg/util/feature
 # k8s.io/component-base v0.20.0 => k8s.io/component-base v0.20.0
 k8s.io/component-base/featuregate
 # k8s.io/klog/v2 v2.4.0
+## explicit
 k8s.io/klog/v2
 # k8s.io/kubernetes v1.20.0
 ## explicit