Skip to content

ZFS Disk Replacement

The process of replacing mirrored zfs disks is fairly simple. The changes are done by zpool attach and detach.

zpool detach <pool> <disk-id>
zpool attach <pool> <disk-id-to-mirror> <disk-id-mirrored-to>

The heavy lifting is done by zfs itself.

process

PREP

  • use the pdf article link to print this before going down
  • If possible pre wipe and check the disks on a separate linux machine (note: /dev/sdf is an placeholder for the disk mounted on that system)

``` root@homebox:~# wipefs -af --backup /dev/sdf /dev/sdf: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49 20 50 41 52 54 /dev/sdf: 8 bytes were erased at offset 0x222ee64e00 (gpt): 45 46 49 20 50 41 52 54 /dev/sdf: 2 bytes were erased at offset 0x000001fe (PMBR): 55 aa /dev/sdc: calling ioctl to re-read partition table: Success root@homebox:~# fdisk /dev/sdf .... Command (m for help): g

Created a new GPT disklabel (GUID: EBC5A0C9-E871-544F-A8EA-E31FCA655F9C).

Command (m for help): w
The partition table has been altered.
Calling ioctl() to re-read partition table.
Syncing disks.

root@homebox:~# badblocks /dev/sdf
....

```

  • insure that you can ssh into the box

On site

The following assumes you have escalated to root privileges (sudo bash), in this case we are replacing /dev/sdc and /dev/sdd in the pool named 'level'

  • check for the correct disk. The following should cause the disk to light up\ ( C when you have identified the disk. Careful with the if/of here).

root@bs2020:~# dd if=/dev/sdc of=/dev/null

  • find the disk in the pool.

``` root@bs2020:~# zpool status pool: devel state: ONLINE scan: resilvered 9.95G in 0h4m with 0 errors on Sat Nov 10 22:00:41 2018 config:

    NAME                        STATE     READ WRITE CKSUM
    devel                       ONLINE       0     0     0
      mirror-0                  ONLINE       0     0     0
        scsi-35000c50054fee503  ONLINE       0     0     0
        scsi-35000c5005501b45b  ONLINE       0     0     0

errors: No known data errors

...
root@bs2020:~# ls -ls /dev/disk/by-id/|grep scsi|grep -v "\-part"
0 lrwxrwxrwx 1 root root  9 Nov 10 21:22 scsi-350000395a8336d34 -> ../../sde
0 lrwxrwxrwx 1 root root  9 Nov 10 21:22 scsi-35000c50054fee503 -> ../../sdd
0 lrwxrwxrwx 1 root root  9 Nov 10 21:56 scsi-35000c5005501b45b -> ../../sdc
0 lrwxrwxrwx 1 root root  9 Nov 10 21:22 scsi-35000cca00b33a264 -> ../../sdf
0 lrwxrwxrwx 1 root root  9 Nov 10 21:22 scsi-3600508e00000000069cf3977618f1408 -> ../../sdg
root@bs2020:~#

``` We notice above that the disk we are looking for is scsi-35000c5005501b45b

  • detach the disk from the pool.

``` root@bs2020:~# zpool detach devel scsi-35000c5005501b45b root@bs2020:~# zpool status pool: devel state: ONLINE scan: resilvered 9.95G in 0h4m with 0 errors on Sat Nov 10 22:00:41 2018 config:

    NAME                      STATE     READ WRITE CKSUM
    devel                     ONLINE       0     0     0
      scsi-35000c50054fee503  ONLINE       0     0     0

errors: No known data errors

  pool: infra
...
root@bs2020:~#

```

  • even if expanding the disk size insure that auto expand is off.

root@bs2020:~# zpool set autoexpand=off devel

  • Swap out the old disk with the new one.

  • find the new disk's id.

root@bs2020:~# partprobe root@bs2020:~# ls -ls /dev/disk/by-id/|grep sdc 0 lrwxrwxrwx 1 root root 9 Nov 10 21:56 scsi-xxxxxxxxxxxxxxx -> ../../sdc ... 0 lrwxrwxrwx 1 root root 9 Nov 10 21:56 xxx-xxxxxxxxxxxxxxx -> ../../sdc

  • If the drive id does not change reboot the server

  • attach the new disk to the zfs pool (scsi-xxxxxxxxxxxxxxxx is the new id from the above step)

root@bs2020:~# zpool attach devel scsi-35000c50054fee503 scsi-xxxxxxxxxxxxxx

  • wait for pool to resliver

```
root@bs2020:~# zpool status pool: devel state: ONLINE status: One or more devices is currently being resilvered. The pool will continue to function, possibly in a degraded state. action: Wait for the resilver to complete. scan: resilver in progress since Sat Nov 10 21:56:04 2018 8.54G scanned out of 9.95G at 35.5M/s, 0h0m to go 8.54G resilvered, 85.85% done config:

    NAME                        STATE     READ WRITE CKSUM
    devel                       ONLINE       0     0     0
      mirror-0                  ONLINE       0     0     0
        scsi-35000c50054fee503  ONLINE       0     0     0
        scsi-35000c5005501b45b  ONLINE       0     0     0  (resilvering)

errors: No known data errors

  pool: infra
...

root@bs2020:~# zpool status
.... repeat until finished reslivering .... 
root@bs2020:~# zpool status
  pool: devel
 state: ONLINE
  scan: scrub repaired 0B in 0h4m with 0 errors on Sat Nov 10 21:58:04 2018
config:

    NAME                        STATE     READ WRITE CKSUM
    devel                       ONLINE       0     0     0
      mirror-0                  ONLINE       0     0     0
        scsi-35000cca00b33a264  ONLINE       0     0     0
        scsi-350000395a8336d34  ONLINE       0     0     0

errors: No known data errors
...

```

  • if expanding disk check for new size and if not expand it

# zfs list and check for larger disk pool

  • repeat process for disk in bay below (we already know its old id from above).

root@bs2020:~# dd if=/dev/sdd of=/dev/null root@bs2020:~# zpool detach devel scsi-35000c50054fee503 ... swap disks ... root@bs2020:~# partprobe root@bs2020:~# ls -ls /dev/disk/by-id/|grep sdd 0 lrwxrwxrwx 1 root root 9 Nov 10 21:56 scsi-yyyyyyyyyyyyyyyy-> ../../sdd ... reboot if necessary ... root@bs2020:~# wipefs -a /dev/sdd ... root@bs2020:~# fdisk /dev/sdd ... root@bs2020:~# zpool attach devel scsi-xxxxxxxxxxxxxx scsi-yyyyyyyyyyyyyyy ... wait for resliver... * use the process below to grow disks to new size

# zpool set autoexpand=on devel # zpool online -e devel scsi-xxxxxxxxxxxxxxxxxxxx # zpool online -e devel scsi-yyyyyyyyyyyyyyyyyyy # zpool set autoexpand=off devel

references

  • https://tomasz.korwel.net/2014/01/03/growing-zfs-pool/
  • https://jsosic.wordpress.com/2013/01/01/expanding-zfs-zpool-raid/
  • https://serverfault.com/questions/5336/how-do-i-make-linux-recognize-a-new-sata-dev-sda-drive-i-hot-swapped-in-without