About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
You added a virtual disk to a running Linux VM in vSphere, but lsblk and fdisk -l do not list it. The guest kernel may simply not have been told to look. Broadcom's instruction is that "To re-scan the SCSI bus entirely, reboot the guest operating system", and if the guest cannot be rebooted, "perform force SCSI rescan" by writing "- - -" to each SCSI host's scan file under /sys/class/scsi_host/. Do that before you touch partitions or /etc/fstab. One scope note: Broadcom's page 324730 is written for volumes attached to a VMware Integrated OpenStack instance, with a sample that rescans PCI first (echo "1" > /sys/bus/pci/rescan); the generic Linux rescan from it is used here.
There are two other situations worth ruling out: a known defect with one old controller type, and a disk that is visible but has no partition or file system yet. Both are covered below.
What the symptom means
A virtual disk attached to a running VM is presented to the guest as a new device on a virtual SCSI (or NVMe) controller. Linux does not always poll for it. In the symptom Broadcom documents, the guest showed only some of the attached volumes until the SCSI hosts were rescanned, and "After running the preceding command, all the volumes should correctly show up in the guest OS".
Broadcom explains the string this way: "the hyphens ""- - -" represent controller channel lun. Therefore, – – – indicates all controllers, all channels, and all luns." Writing it to a host's scan file asks that SCSI host to look for everything attached to it.
A second Broadcom article, for a hot-added LSI Logic Parallel controller and disk, gives a two-line rescan that covers every rescan and scan file under /sys, and says the cause "is caused by a limitation in Linux kernel". That Broadcom page lists vSphere 6.x and 7.0 and Oracle Linux 7.6 and 8.0 and Debian 9 and 10 guests, so treat it as a pointer to the method rather than a statement about current releases.
Broadcom also documents a defect, not a design behaviour: on ESXi 5.0 to 6.5, a disk hot-added to an existing LSI Logic SAS controller on a Linux kernel 4.2 or newer "cannot be detected" even by rescan-scsi-bus.sh. It "is resolved in ESXi 6.5 U2", because Broadcom's page says the issue is resolved in ESXi 6.5 U2.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| The kernel has not rescanned the SCSI host | lsblk has no new row; the VM settings in vSphere do show the disk |
Rescan and run lsblk again |
| The disk is on a newly added controller | Neither the controller nor its disk is listed | Check the VM settings for a new controller; rescan; Broadcom's LSI Logic Parallel article describes this case |
| Defect with a hot-added disk on an existing LSI Logic SAS controller (ESXi 5.0 to 6.5) | Rescans do not help; a reboot does | Controller type in the VM settings; ESXi version older than 6.5 U2 |
| The disk is visible but empty | It appears in lsblk with no partitions and no file system |
lsblk -f shows an empty file system column |
| The disk is not really attached to this VM | Not in the vSphere VM settings, or on another VM | Check the VM's Edit Settings in vSphere |
| The disk appears, but the next boot stops on a missing mount | A wait or emergency shell after you added an /etc/fstab line |
The fstab line references a device by /dev/sdX name, or lacks nofail |
Prove the cause before changing anything
Start with what vSphere says, because a guest cannot show a disk that is not attached. Open the VM's settings and note the disk, its size and its controller. Then read, inside the guest:
lsblk
lsblk -f
ls /sys/class/scsi_host
dmesg -T
lsblk lists block devices. The -f option "output[s] info about filesystems" and is equivalent to showing the name, file system type, label, UUID, free space and mount points, so an empty disk shows a name and size but nothing else. ls /sys/class/scsi_host lists the SCSI host entries (named host0, host1 and so on) that Broadcom's command refers to. dmesg -T prints the kernel log with human-readable timestamps; the manual page warns that "the timestamp could be inaccurate", so use it to find messages near the time of the change, not for exact timing. A new disk can produce lines about a SCSI disk being attached after a rescan.
A good result is a new row in lsblk of the size you added. A bad result, after the rescan, is still no row and no kernel message.
Example (illustrative): app01 has sda (40 GiB) and a controller host0. You add a 100 GiB disk in vSphere with the VM running. lsblk shows only sda. You list /sys/class/scsi_host, see host0 and host1, and write "- - -" to each host's scan file. lsblk now shows sdb at 100G with no partitions, and lsblk -f shows an empty file system column. The disk was always attached; the kernel only needed to scan. The next job is to create a file system and a mount, and the 40 GiB sda is untouched.
Fix
Fix 1: rescan the SCSI hosts
For each name listed by ls /sys/class/scsi_host, as root:
echo "- - -" > /sys/class/scsi_host/host0/scan
Repeat with host1 and the others. Broadcom's article writes the path with a * in place of the host name, so that one command reaches every host. Broadcom's other article gives this rescan, which covers every rescan or scan file:
for i in `find /sys/ -iname rescan`;do echo 1 >$i; done;
for i in `find /sys/ -iname scan`;do echo "- - -" >$i; done;
Risk: the loop writes to every file called rescan or scan anywhere under /sys, which is broader than needed and also asks every attached device to rescan, including disks that are in use. It is Broadcom's published command, not a minimal one. Prefer the per-host command. Broadcom documents no side effects and no undo for either form, and the sources do not say what a rescan does to a busy production disk, so choose a quiet period.
Run lsblk again. If the disk appears, stop here and go to the file system step.
Fix 2: reboot the guest
If the rescan does not show the disk, Broadcom's first instruction applies: "To re-scan the SCSI bus entirely, reboot the guest operating system."
Disruptive step: a reboot interrupts every service on the VM. It does not change the attached disk. If you add the disk's mount to /etc/fstab before the reboot, follow Fix 4 first, or the reboot may stop at a missing device.
Fix 3: the LSI Logic SAS defect on old hosts
On ESXi 5.0 to 6.5 before 6.5 U2, Broadcom's workaround is to reboot the VM after hot-adding, or to add the disk together with a new LSI Logic SAS controller, "then the new disk can be detected by the guest." Broadcom's page says the issue is resolved in ESXi 6.5 U2.
Fix 4: file system and mount
When lsblk shows the disk, the work is the same as on a physical server: create a partition and file system with your distribution's tools, then mount it. The steps here stop at identification. The part that goes wrong at the next boot is the mount entry.
- Identify the file system by UUID.
lsblk -fprints it, andblkidis described as the authoritative tool for file system information.fstab(5)says aUUID=<uuid>may be given instead of a device, because/dev/sdbcan change name between boots. - Add
nofail. Thefstab(5)manual saysnofailmeans "do not report errors for this device if it does not exist." In systemd terms, anofailmount "will be only wanted, not required, by local-fs.target". - Decide how long to wait.
x-systemd.device-timeout=configures "how long systemd should wait for a device to show up before giving up on an entry from /etc/fstab", andDefaultDeviceTimeoutSec="Defaults to 90 s".
An entry in this form uses all three:
UUID=0a1b2c3d-1111-2222-3333-444455556666 /data ext4 defaults,nofail,x-systemd.device-timeout=10s 0 2
Before rebooting, findmnt --verify checks that /etc/fstab can be parsed and used.
Risk: editing /etc/fstab can stop the next boot. Copy the file first and restore the copy if the check reports an error. See "Linux VM in Emergency Mode After a Disk Change: fstab Checks" if a boot has already stopped.
Confirm the fix
lsblklists the new disk at the size you added in vSphere. Compare the size in GiB with the vSphere setting.lsblk -fshows the file system type, UUID and mount point once you have created and mounted it.findmnt --verifyreports no problem with/etc/fstab.- After the next planned reboot, the mount is present. Check it, and look at
systemctl --failedif it is not.
What to watch afterwards
- Device names. A disk that was
sdbtoday may be another name after a change of controller or order. Mount by UUID. - Boot time. A mount without
nofailmakes the boot wait for the device, and the default wait is 90 seconds. See "Linux boot stuck on "A start job is running for ...": what systemd is waiting for" for how to read those waits. - Controller type. If you add disks to a running VM regularly, the controller type matters. Keep a note of which controller each disk sits on.
- Size changes. A disk that appears at the old size after an extend is a different fault; see "VMware disk made bigger but Linux still shows the old size: rescan, growpart, pvresize and lvextend in order".
Read next
Related articles you may want next.
- TroubleshootingLinux VM in Emergency Mode After a Disk Change: fstab ChecksWhy a systemd Linux VM stops at an emergency shell after a disk change, how to find the failing fstab entry read-only, and when nofail is the right choice.Read article
- VMwareVMware disk made bigger but Linux still shows the old size: rescan, growpart, pvresize and lvextend in orderVirtual disk grown in vSphere but Linux shows the old size? Rescan, then growpart, pvresize, lvextend and resize2fs or xfs_growfs, with checks.Read article
- HostingAdding CPU, memory or disk to a hosted server: which changes need a restartPlan CPU, memory and disk upgrades for hosted servers. Check which changes support live expansion, the conditions that apply and when a planned restart is needed.Read article
Risks and limits
- These commands have not been run or lab-tested. They come from the Broadcom articles and manual pages listed below.
- The Broadcom articles are older than current ESXi and guest releases. The LSI Logic articles name specific ESXi and kernel versions and may not apply to you.
- The sources do not describe NVMe or paravirtual (PVSCSI) controllers hot-adding a disk to Linux, so no claim is made about them beyond the generic rescan.
- Writing to files under
/sysas root affects the running kernel. Check the host name before you write. - Partition and file system creation, and the choice of file system, are outside what the sources cover.
Sources
- All volumes attached to an instance are not shown in a Linux guest OS (KB 324730) — Broadcom, updated 11-13-2026, accessed 18 November 2026
- Hot added LSI Logic Parallel SCSI controller and disk not recognized in Linux (KB 344007) — Broadcom, updated 06-03-2026, accessed 18 November 2026
- Hot-added new hard disk attached to an existing LSI Logic SAS controller cannot be detected in Linux kernel 4.2 or newer (KB 320237) — Broadcom, updated 11-04-2024, accessed 18 November 2026
- systemd.mount(5) — man7.org, no date shown, accessed 18 November 2026
- systemd-system.conf(5) — man7.org, no date shown, accessed 18 November 2026
- fstab(5) — man7.org, no date shown, accessed 18 November 2026
- lsblk(8), blkid(8), findmnt(8), dmesg(1) — Debian manpages, util-linux 2.38.1-5, no date shown, accessed 18 November 2026
This article was originally published on .