drivelist is a tool to make managing large numbers of drives on Linux systems easier. On my system, it's able to walk through roughly 100 drives in about a half second and identify how they're used.
- It can enumerate all drives and show device type, serial number, WWN, enclosure, and bay for each, making it easier to physically locate drives.
- It can identify which drives are in use and easily show unused drives (via the --unused flag)
- It can show empty/unused enclosure bays, which may help spot dead drives.
You'll need to have a recent Go compiler installed. This also depends
on having libzfs installed; on Ubuntu you can install this via apt install libzfslinux-dev.
Then do something like this:
$ go checkout https://github.com/scottlaird/drivelist.git
$ go get
$ go build cmd/drivelist/drivelist.go
This will leave a runable drivelist binary in the current directory.
drivelist currently only supports SCSI-like drives on Linux. SAS, SATA, and NVMe are fine. Actual SCSI drives might work, if you could find one that actually still works. USB probably works, but if you have a dozen or more USB drives on one system then you have a different problem.
Currently, drivelist can identify drives in use by checking mountpoints and by looking into ZFS pools and vdevs. It successfully deals with ZFS spares, log devices, pending resilvers, and so on.
Linux MD, LVM, and btrfs support is currently missing. Stub code exists for MD and LVM, but I'm not currently using either.
By default, running drivelist will show all devices in the current system along with a bit of data about each.
Example:
$ ./drivelist
Device Name Model WWN Serial Expander Bay Size
=========== ===== === ====== ======== === ====
sda HUH721008AL5204 0x5000cca25206c808 7SG3RM2G expander-4:0 0 8 TB
sdaa H7280A520SUN8.0T 0x5000cca2548eecec 001619PJLREV_VKJJLREV expander-11:0 54 8 TB
sdab HUH72808CLAR8000 0x5000cca26103eb9c VJG24UZX expander-11:0 10 8 TB
sdac PX02SMQ160 0x500003964c8806c4 5520A0C2T2AA expander-11:0 33 1600 GB
sdad HUH72808CLAR8000 0x5000cca261035808 VJG1V09X expander-11:0 20 8 TB
...
The drivelist tool collects more data than can reasonably be shown on one line in a terminal; adding the --allfields flag will show all fields.
$ ./drivelist --allfields
Device Name Devices WWN Sys Path Model Serial Uses Generic Device Expander Expander Path Bay Size
=========== ======= === ======== ===== ====== ==== ============== ======== ============= === ====
sda /dev/sda,/dev/disk/by-id/scsi-SHGST_HUH721008AL5204_7SG3RM2G,/dev/disk/by-path/pci-0000:01:00.0-sas-exp0x500304800000007f-phy8-lun-0,/dev/disk/by-vdev/Ab0,/dev/disk/by-id/wwn-0x5000cca25206c808,/dev/disk/by-id/scsi-35000cca25206c808 0x5000cca25206c808 /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/host4/port-4:0/expander-4:0/port-4:0:0/end_device-4:0:0/target4:0:0/4:0:0:0/block/sda HUH721008AL5204 7SG3RM2G zfs > space 5925914041408872576 > raidz2 1731578064309998343 > disk 1837064446111292648 /dev/bsg/end_device-4:0:0 expander-4:0 //sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/host4/port-4:0/expander-4:0 0 8 TB
sdaa /dev/sdaa,/dev/disk/by-id/scsi-SHGST_H7280A520SUN8.0T_001619PJLREV_VKJJLREV,/dev/disk/by-path/pci-0000:03:00.0-sas-exp0x5000ccab0200947e-phy12-lun-0,/dev/disk/by-vdev/D54,/dev/disk/by-id/wwn-0x5000cca2548eecec,/dev/disk/by-id/scsi-35000cca2548eecec 0x5000cca2548eecec /sys/devices/pci0000:00/0000:00:03.0/0000:03:00.0/host11/port-11:0/expander-11:0/port-11:0:6/end_device-11:0:6/target11:0:13/11:0:13:0/block/sdaa H7280A520SUN8.0T 001619PJLREV_VKJJLREV zfs > space 5925914041408872576 > raidz2 5236016460003016805 > disk 9810514795403010748 /dev/bsg/end_device-11:0:6 expander-11:0 //sys/devices/pci0000:00/0000:00:03.0/0000:03:00.0/host11/port-11:0/expander-11:0 54 8 TB
sdab /dev/sdab,/dev/disk/by-id/scsi-35000cca26103eb9c,/dev/disk/by-vdev/D10,/dev/disk/by-id/scsi-SHITACHI_HUH72808CLAR8000_VJG24UZX,/dev/disk/by-path/pci-0000:03:00.0-sas-exp0x5000ccab0200947e-phy14-lun-0,/dev/disk/by-id/wwn-0x5000cca26103eb9c 0x5000cca26103eb9c /sys/devices/pci0000:00/0000:00:03.0/0000:03:00.0/host11/port-11:0/expander-11:0/port-11:0:7/end_device-11:0:7/target11:0:14/11:0:14:0/block/sdab HUH72808CLAR8000 VJG24UZX zfs > space 5925914041408872576 > raidz2 12372305547527317295 > disk 4484622911110778645 /dev/bsg/end_device-11:0:7 expander-11:0 //sys/devices/pci0000:00/0000:00:03.0/0000:03:00.0/host11/port-11:0/expander-11:0 10 8 TB
...
There is also a --fields= flag for picking specific fields:
$ ./drivelist --fields=devicename,genericdevice,size
Device Name Generic Device Size
=========== ============== ====
sda /dev/bsg/end_device-4:0:0 8 TB
sdaa /dev/bsg/end_device-11:0:6 8 TB
sdab /dev/bsg/end_device-11:0:7 8 TB
sdac /dev/bsg/end_device-11:0:8 1600 GB
sdad /dev/bsg/end_device-11:0:9 8 TB
...
Finally, there is an --unused flag that only shows devices with no
known use in the system. This filters out drives that have mounted
filesystems as well as drives that are part of ZFS pools, leaving only
unused devices. This is very useful with large numbers of drives, as
it's easy to lose track of physically present drives during
migrations, etc, and end up with installed and powered-on but unused
disks in a system with many drives.
It's also useful for identifying failed devices that have been removed from ZFS but not yet physically removed from the system.
$ ./drivelist --unused
Device Name Model WWN Serial Expander Bay Size
=========== ===== === ====== ======== === ====
sdal PX02SMU020 0x500003964c8806e4 5520A0CAT2AA expander-11:0 30
sdbt HUH728080ALE601 0x5000cca260c165e2 VLG32AEY expander-11:1 17 8 TB
- Improve testing.
- Add a
testdata/directory with multiple sets of test/sysand/procdata, and run tests across each. - Find a reasonable way to fake zfs.PoolOpenAll().
- Add a
- Flesh out MD support
- Flesh out LVM support
- Add btrfs support
- Verify/fix USB support
- Improve filtering; make it possible to show all devices, empty drive bays, or any reasonable subset.
- Add CSV output, simple tab-delmited output, and make headers optional.
- Add sorting
- Consider adding state, ideally shared over the network across
multiple servers.
- Mark disks as bad and see them flagged wherever they appear.
- Answer "when did this disk first appear?" and other lifecycle questions.
- Track transient issues, past uses, etc, as make sense.
- Add better SAS support, to identify link paths, speeds, and widths.
- Add support for managing SAS enclosure LEDs (via lcdctl or similar?) to flag specific drives for removal, etc.