fix: add cooldown to health monitor unmount to prevent log spam

After force-unmounting a broken staging path, suppress repeated attempts
for the same path for 5 minutes. This prevents noisy warning logs when
kubelet is slow to re-stage (e.g., deep CrashLoopBackOff backoff).

Author: fenio

pkg/driver/node.go

@@ -7,6 +7,7 @@ import (
 	"path/filepath"
 	"strings"
 	"syscall"
+	"time"
 
 	"github.com/container-storage-interface/spec/lib/go/csi"
 	"github.com/nasty-project/nasty-csi/pkg/metrics"
@@ -43,6 +44,7 @@ type NodeService struct {
 	nodeRegistry    *NodeRegistry
 	nvmeConnectSem  chan struct{}
 	stopCh          chan struct{}
+	recentUnmounts  map[string]time.Time // tracks recently unmounted staging paths to avoid log spam
 	nodeID          string
 	testMode        bool
 	enableDiscovery bool
@@ -61,6 +63,7 @@ func NewNodeService(nodeID string, apiClient nastyapi.ClientInterface, testMode
 		enableDiscovery: enableDiscovery,
 		nvmeConnectSem:  make(chan struct{}, maxConcurrentNVMeConnects),
 		stopCh:          make(chan struct{}),
+		recentUnmounts:  make(map[string]time.Time),
 	}
 }
 

pkg/driver/node_recovery.go

@@ -81,6 +81,10 @@ func (s *NodeService) runHealthCheck() {
 	s.unmountBrokenStagingPaths(ctx)
 }
 
+// unmountCooldown is how long we suppress repeated unmount attempts for the
+// same staging path. This prevents log spam when kubelet is slow to re-stage.
+const unmountCooldown = 5 * time.Minute
+
 // unmountBrokenStagingPaths finds staging mounts where the source block device
 // no longer exists and force-unmounts them so kubelet can re-stage.
 func (s *NodeService) unmountBrokenStagingPaths(ctx context.Context) {
@@ -95,6 +99,14 @@ func (s *NodeService) unmountBrokenStagingPaths(ctx context.Context) {
 		return
 	}
 
+	// Expire old entries from the cooldown map
+	now := time.Now()
+	for path, ts := range s.recentUnmounts {
+		if now.Sub(ts) > unmountCooldown {
+			delete(s.recentUnmounts, path)
+		}
+	}
+
 	// Filter for our CSI driver's staging paths
 	const stagingPathMarker = "/plugins/kubernetes.io/csi/nasty.csi.io/"
 	for _, line := range strings.Split(string(output), "\n") {
@@ -110,6 +122,11 @@ func (s *NodeService) unmountBrokenStagingPaths(ctx context.Context) {
 		source := fields[0]
 		target := fields[1]
 
+		// Skip if we recently unmounted this path (avoid log spam)
+		if _, cooldown := s.recentUnmounts[target]; cooldown {
+			continue
+		}
+
 		// Check if the source device still exists
 		if _, statErr := os.Stat(source); statErr == nil {
 			// Device exists — check if it's actually accessible
@@ -132,6 +149,9 @@ func (s *NodeService) unmountBrokenStagingPaths(ctx context.Context) {
 			klog.Infof("Successfully unmounted broken staging path %s — kubelet will re-stage", target)
 		}
 		umountCancel()
+
+		// Record this unmount to suppress repeated attempts
+		s.recentUnmounts[target] = now
 	}
 }