lima-vm/lima · error

failed to run attach disk %#q, in use by instance %#q

Error message

failed to run attach disk %#q, in use by instance %#q

What it means

Cmdline checks each additional disk's `disk.Instance` field; if it is non-empty the disk is already attached to another running Lima instance, so krunkit refuses to attach it a second time and returns `failed to run attach disk %#q, in use by instance <name>`. Lima disks are single-attach: locking to more than one instance would corrupt data.

Source

Thrown at pkg/driver/krunkit/krunkit_darwin_arm64.go:69

		"--device", fmt.Sprintf("virtio-vsock,port=%d,socketURL=%s,connect", vSockPort, filepath.Join(inst.Dir, filenames.GuestAgentSock)),
	}

	if inst.Config.SSH.OverVsock != nil && *inst.Config.SSH.OverVsock {
		sshVsockPath := filepath.Join(inst.Dir, sshVsockSock)
		args = append(args, "--device", fmt.Sprintf("virtio-vsock,port=22,socketURL=%s,connect", sshVsockPath))
	}

	// Add additional disks
	if len(inst.Config.AdditionalDisks) > 0 {
		ctx := context.Background()
		diskUtil := proxyimgutil.NewDiskUtil(ctx)
		for _, d := range inst.Config.AdditionalDisks {
			disk, derr := store.InspectDisk(d.Name, d.FSType)
			if derr != nil {
				return nil, fmt.Errorf("failed to load disk %#q: %w", d.Name, derr)
			}
			if disk.Instance != "" {
				return nil, fmt.Errorf("failed to run attach disk %#q, in use by instance %#q", disk.Name, disk.Instance)
			}
			if lerr := disk.Lock(inst.Dir); lerr != nil {
				return nil, fmt.Errorf("failed to lock disk %#q: %w", d.Name, lerr)
			}
			extraDiskPath := filepath.Join(disk.Dir, filenames.DataDisk)
			logrus.Infof("Mounting disk %#q on %#q", disk.Name, disk.MountPoint)
			if cerr := diskUtil.Convert(ctx, raw.Type, extraDiskPath, extraDiskPath, nil, true); cerr != nil {
				return nil, fmt.Errorf("failed to convert extra disk %#q to raw: %w", extraDiskPath, cerr)
			}
			args = append(args, "--device", fmt.Sprintf("virtio-blk,path=%s,format=raw", extraDiskPath))
		}
	}

	// Network commands
	networkArgs, err := buildNetworkArgs(inst)
	if err != nil {
		return nil, fmt.Errorf("failed to build network arguments: %w", err)
	}

View on GitHub (pinned to dd909d0973)

Solutions

1. Stop/ungister the instance currently holding the disk (`limactl stop <other-instance>`), then start yours.
2. If the holder instance is stale/gone, remove the stale reference: `limactl disk undeploy <disk>` (or delete disk metadata) so it can be re-attached.
3. Give each instance its own disk: create a new disk with `limactl disk create` and update `additionalDisks`.
4. Run `limactl disk list` to see which instance occupies the disk before starting.

Example fix

# before (second instance lima.yaml)
additionalDisks:
  - name: shared-data   # attached to instance "dev"
# after
additionalDisks:
  - name: shared-data-2 # limactl disk create shared-data-2 --size 50G

// Go: check the disk is free before start
if diskJSON, err := exec.Command("limactl", "disk", "list", diskName).Output(); err == nil {
    if bytes.Contains(diskJSON, []byte("instance:")) && !bytes.Contains(diskJSON, []byte("instance: \"\"")) {
        return fmt.Errorf("disk %q in use; stop the holder instance first", diskName)
    }
}

if err := startInstance(ctx, name); err != nil {
    if strings.Contains(err.Error(), "in use by instance") {
        holder := extractInstanceName(err.Error())
        _ = exec.Command("limactl", "stop", holder).Run()
        return startInstance(ctx, name) // retry once
    }
    return err
}

When it happens

Trigger: Starting an instance whose `additionalDisks` includes a disk whose metadata records it as in use by a different (usually still registered as running) instance; happens when a previous instance crashed without releasing the disk or two instances reference the same disk.

Common situations: Cloning an instance config that shares `additionalDisks` between instances; a previous VM crashed leaving the disk lock/instance field stale; trying to boot the same disk in a second instance for parallel workloads.

Related errors

• failed to lock disk %#q: %w
• failed to load disk %#q: %w
• failed to convert extra disk %#q to raw: %w
• unimplemented by the krunkit driver
• failed to build network arguments: %w

AI-assisted analysis of lima-vm/lima@dd909d0973 (2026-09-01). Data as JSON: /api/errors/cc4def4bc9cb9c59. Report an issue: GitHub.