Wandering Thoughts archives

Using lsblk to get extremely useful information about disks

Every so often I need to know the serial number of a disk, generally because it's the only way to identify one particular disk out of two (or more) identical ones. As one example, perhaps I need to replace a failed drive that's one of a pair. You can get this information from the disks through smartctl, but the process is somewhat annoying if you just want the serial number, especially if you want it for multiple disks.

(Sometimes you have a dead disk so you need to find it by process of elimination starting from the serial numbers of all of the live disks.)

I've used lsblk for some time to get disk UUIDs and raid UUIDs, but I never looked very deeply at its other options. Recently I discovered that lsblk can do a lot more, and in particular it can report disk serial numbers (as well as a bunch of other handy information) in an extremely convenient form. It's simplest to just show you an example:

$ lsblk -o NAME,SERIAL,HCTL,TRAN,MODEL --nodeps /dev/sd?
NAME SERIAL          HCTL       TRAN   MODEL
sda  S21NNXCGAxxxxxH 0:0:0:0    sata   Samsung SSD 850 
sdb  S21NNXCGAxxxxxE 1:0:0:0    sata   Samsung SSD 850 
sdc  Zxxxxx4E        2:0:0:0    sata   ST500DM002-1BC14
sdd  WD-WMC5K0Dxxxxx 4:0:0:0    sata   WDC WD1002F9YZ-0
sde  WD-WMC5K0Dxxxxx 5:0:0:0    sata   WDC WD1002F9YZ-0

(For obscure reasons I don't feel like publishing the full serial numbers of our disks. It might be harmless to do so, but let's not find out otherwise the hard way.)

You can get a full list of possible fields with 'lsblk --help', along with generally what they mean, although you'll find that some of them are less useful than you might guess. VENDOR is always 'ATA' for me, for example, and KNAME is the same as NAME for my systems; TRAN is usually 'sata', as here, but we have some machines where it's different. Looking for a PHY-SEC that's not 512 is a convenient way to find advanced format drives, which may be surprisingly uncommon in some environments. SIZE is another surprisingly handy field; if you know you're looking for a disk of a specific size, it lets you filter disks in and out without checking serial numbers or even the specific model, if you have multiple different sized drives from one vendor such as WD or Seagate.

(--nodeps tells lsblk to just report on the devices that you gave it and not also include their partitions, software RAID devices that use them, and so on.)

This compact lsblk output is great for summarizing all of the disks on a machine in something that's easy to print out and use. Pretty much everything I need to know is one spot and I can easily use this to identify specific drives. I'm quite happy to have stumbled over this additional use of lsblk, and I plan to make much more use of it in the future. Possibly I should routinely collect this output for my machines and save it away.

(This entry is partly to write down the list of lsblk fields that I find useful so I don't have to keep remembering them or sorting through lsblk --help and trying to remember the fields that are less useful than they sound.)