fix(disks): Remove unmounting from longhorn cleanup. It is done elsewhere.

.github/workflows/run-tests.yml

@@ -16,7 +16,7 @@ jobs:
         run: devbox run go test -v ./pkg/...
 
       - name: Run integration tests
-        run: devbox run go test -v ./tests/integration
+        run: dist/bloom test tests/integration/step/*/*/bloom.yaml
 
       - name: Run UI tests
         run: |

pkg/args/args.go

@@ -9,6 +9,7 @@ import (
 	"regexp"
 	"strings"
 
+	"github.com/silogen/cluster-bloom/pkg/mockablecmd"
 	log "github.com/sirupsen/logrus"
 	"github.com/spf13/viper"
 	"gopkg.in/yaml.v3"
@@ -202,6 +203,46 @@ func ValidateSkipDiskCheckConsistency(skipDiskCheckStr string) error {
 	return nil
 }
 
+// validatePremountedNotBloomManaged checks that CLUSTER_PREMOUNTED_DISKS paths are not bloom-managed in /etc/fstab
+func validatePremountedNotBloomManaged(diskList []string) error {
+	const bloomFstabTag = "# managed by cluster-bloom"
+
+	fstabContent, err := mockablecmd.ReadFile("ValidateArgs.ReadFstab", "/etc/fstab")
+	if err != nil {
+		// If we can't read fstab, warn but don't fail validation
+		log.Warnf("Could not read /etc/fstab to validate CLUSTER_PREMOUNTED_DISKS: %v", err)
+		return nil
+	}
+
+	lines := strings.Split(string(fstabContent), "\n")
+	for lineNum, line := range lines {
+		trimmedLine := strings.TrimSpace(line)
+
+		// Skip lines that aren't bloom-managed
+		if !strings.HasSuffix(trimmedLine, bloomFstabTag) {
+			continue
+		}
+
+		// Extract mount point from bloom-managed entry
+		fields := strings.Fields(trimmedLine)
+		if len(fields) < 2 {
+			continue
+		}
+
+		mountPoint := fields[1]
+
+		// Check if this mount point is in CLUSTER_PREMOUNTED_DISKS
+		for _, disk := range diskList {
+			disk = strings.TrimSpace(disk)
+			if disk == mountPoint {
+				return fmt.Errorf("CLUSTER_PREMOUNTED_DISKS contains %s which is tagged '# managed by cluster-bloom' in /etc/fstab at line %d:\n  %s\nPlease delete the tag from the fstab line first or use a different mount point", disk, lineNum+1, trimmedLine)
+			}
+		}
+	}
+
+	return nil
+}
+
 // ValidateLonghornDisksArg validates CLUSTER_PREMOUNTED_DISKS configuration
 func ValidateLonghornDisksArg(disks string) error {
 	selectedDisks := viper.GetString("CLUSTER_DISKS")
@@ -222,6 +263,11 @@ func ValidateLonghornDisksArg(disks string) error {
 				return fmt.Errorf("CLUSTER_PREMOUNTED_DISKS contains a path that does not exist: %s", disk)
 			}
 		}
+
+		// Check that none of the CLUSTER_PREMOUNTED_DISKS are bloom-managed in /etc/fstab
+		if err := validatePremountedNotBloomManaged(diskList); err != nil {
+			return err
+		}
 	}
 
 	// Both cannot be set

pkg/steps.go

@@ -225,7 +225,7 @@ var SetupAndCheckRocmStep = Step{
 				if strings.HasPrefix(trimmedLine, "WARNING:") {
 					continue
 				}
-				
+
 				parts := strings.Fields(line)
 				if len(parts) > 0 {
 					if _, err := strconv.Atoi(parts[0]); err != nil {
@@ -238,7 +238,7 @@ var SetupAndCheckRocmStep = Step{
 				}
 			}
 		}
-		
+
 		if !validLineFound {
 			LogMessage(Error, "rocm-smi did not return any valid GPU lines: "+string(output))
 			return StepResult{
@@ -1111,20 +1111,6 @@ var CleanLonghornMountsStep = Step{
 				return stepResult
 			}
 
-			// Find /mnt/disk* mount points that contain longhorn-disk.cfg and unmount them
-			shellCmd := `
-				for mount_point in /mnt/disk*; do
-					if [ -d "$mount_point" ] && find "$mount_point" -name "longhorn-disk.cfg" 2>/dev/null | grep -q .; then
-						echo "Found longhorn-disk.cfg in $mount_point, unmounting..."
-						sudo umount -lf "$mount_point" 2>/dev/null || true
-					fi
-				done
-			`
-			stepResult = shellCmdHelper(shellCmd)
-			if stepResult.Error != nil {
-				return stepResult
-			}
-
 			// Find and unmount CSI volume mounts
 			stepResult = shellCmdHelper("sudo umount -Af /var/lib/kubelet/pods/*/volumes/kubernetes.io~csi/pvc-* 2>/dev/null || true")
 			if stepResult.Error != nil {

tests/integration/step/01_ValidateArgsStep/01-valid-first-node/bloom.yaml

@@ -0,0 +1,10 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+GPU_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_DISKS: /dev/nvme0n1
+RKE2_INSTALLATION_URL: https://get.rke2.io
+ROCM_BASE_URL: https://repo.radeon.com/amdgpu-install/7.0.2/ubuntu/
+ROCM_DEB_PACKAGE: amdgpu-install_7.0.2.70002-1_all.deb

tests/integration/step/01_ValidateArgsStep/02-valid-additional-node/bloom.yaml

@@ -0,0 +1,7 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: false
+GPU_NODE: false
+SERVER_IP: 192.168.1.100
+JOIN_TOKEN: K107c7e3e6e3e6e3e6e3e6e3e6e3e6e3e6e3e6::server:abc123def456
+CLUSTER_DISKS: /dev/nvme0n1
+RKE2_INSTALLATION_URL: https://get.rke2.io

tests/integration/step/01_ValidateArgsStep/03-error-both-disks-set/bloom.yaml

@@ -0,0 +1,9 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_DISKS: /dev/nvme0n1
+CLUSTER_PREMOUNTED_DISKS: /tmp
+
+expected_error: "CLUSTER_PREMOUNTED_DISKS and CLUSTER_DISKS cannot both be set"

tests/integration/step/01_ValidateArgsStep/04-error-neither-disks-set/bloom.yaml

@@ -0,0 +1,8 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+NO_DISKS_FOR_CLUSTER: false
+
+expected_error: "either CLUSTER_PREMOUNTED_DISKS or CLUSTER_DISKS must be set"

tests/integration/step/01_ValidateArgsStep/05-error-premounted-not-absolute/bloom.yaml

@@ -0,0 +1,8 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_PREMOUNTED_DISKS: mnt/disk0
+
+expected_error: "CLUSTER_PREMOUNTED_DISKS contains a non-absolute path: mnt/disk0"

tests/integration/step/01_ValidateArgsStep/06-error-premounted-not-exists/bloom.yaml

@@ -0,0 +1,8 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_PREMOUNTED_DISKS: /mnt/nonexistent
+
+expected_error: "CLUSTER_PREMOUNTED_DISKS contains a path that does not exist: /mnt/nonexistent"

tests/integration/step/01_ValidateArgsStep/07-error-premounted-bloom-tagged/bloom.yaml

@@ -0,0 +1,17 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_PREMOUNTED_DISKS: /tmp
+
+mocks:
+  # Mock /etc/fstab with bloom-tagged entry
+  ValidateArgs.ReadFstab:
+    output: |
+      # /etc/fstab: static file system information.
+      UUID=12345 / ext4 errors=remount-ro 0 1
+      UUID=abc123 /tmp ext4 defaults,nofail 0 2 # managed by cluster-bloom
+      UUID=def456 /mnt/disk1 ext4 defaults,nofail 0 2
+
+expected_error: "CLUSTER_PREMOUNTED_DISKS contains /tmp which is tagged '# managed by cluster-bloom' in /etc/fstab at line 3"

tests/integration/step/01_ValidateArgsStep/08-error-both-disabled-and-enabled/bloom.yaml

@@ -0,0 +1,9 @@
+ENABLED_STEPS: ValidateArgsStep
+DISABLED_STEPS: SetupRKE2Step
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_DISKS: /dev/nvme0n1
+
+expected_error: "DISABLED_STEPS and ENABLED_STEPS cannot both be set"

tests/integration/step/01_ValidateArgsStep/09-error-invalid-ip/bloom.yaml

@@ -0,0 +1,7 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: false
+SERVER_IP: not-an-ip
+JOIN_TOKEN: K107c7e3e6e3e6e3e6e3e6e3e6e3e6e3e6e3e6::server:abc123def456
+CLUSTER_DISKS: /dev/nvme0n1
+
+expected_error: "invalid IP address"

tests/integration/step/01_ValidateArgsStep/10-error-short-join-token/bloom.yaml

@@ -0,0 +1,7 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: false
+SERVER_IP: 192.168.1.100
+JOIN_TOKEN: tooshort
+CLUSTER_DISKS: /dev/nvme0n1
+
+expected_error: "JOIN_TOKEN is too short"

tests/integration/step/01_ValidateArgsStep/11-valid-premounted-clean-fstab/bloom.yaml

@@ -0,0 +1,15 @@
+ENABLED_STEPS: ValidateArgsStep
+FIRST_NODE: true
+DOMAIN: example.com
+USE_CERT_MANAGER: false
+CERT_OPTION: generate
+CLUSTER_PREMOUNTED_DISKS: /tmp
+
+mocks:
+  # Mock /etc/fstab WITHOUT bloom-tagged entries for /tmp
+  ValidateArgs.ReadFstab:
+    output: |
+      # /etc/fstab: static file system information.
+      UUID=12345 / ext4 errors=remount-ro 0 1
+      UUID=abc123 /tmp ext4 defaults,nofail 0 2
+      UUID=def456 /mnt/disk1 ext4 defaults,nofail 0 2 # managed by cluster-bloom