Possible Error In Proxmox ZFS Wiki?

[mhayhurst]

Hello Everyone,

In the ZFS: Tips and Tricks Wiki about 3/4 the way down discussing copying the partition table, it appears the syntax for the sgdisk command is wrong:

root@pve:~# sgdisk --replicate=/dev/sdb /dev/sda

/dev/sdb is the failed disk and in this example we are suppose to be copying the /dev/sda partition table to /dev/sdb. Am I correct or has my coffee not kicked in yet?

 

[fabian]

no, it is correct although a more readable variant might be

sgdisk /dev/source --replicate /dev/target

to keep with the established source first, target second semantics of most UNIX commands.. getopts and thus sgdisk does not care about the order..

 

[GadgetPig]

I had the same thoughts when I first read these steps.

I found some info here:

https://askubuntu.com/questions/579...rtition-scheme-from-one-hard-drive-to-another

From what I understand, bootable ZFS pool uses only 2 disks? So you copy from good boot disk to new target disk:

sgdisk --replicate=/dev/target /dev/source
or
sgdisk -R/dev/target /dev/source

And then sgdisk -G /dev/sdb (randomize the disk's GUID and all partitions' unique GUIDs)

They mentioned another method which I really liked:

sgdisk --backup=table /dev/sda (backup table from good source ZFS boot disk)
sgdisk --load-backup=table /dev/sdb (load table to new target ZFS boot disk)
sgdisk -G /dev/sdb (randomize target ZFS boot disk's GUID and all partitions' unique GUIDs)

If someone can confirm this looks right, I had WAY too much coffee this AM

[EDIT]
Thank You Fabian for your post, it makes much more sense:
"sgdisk /dev/source --replicate /dev/target"

 

[CharlesErickT]

From what I understand, bootable ZFS pool uses only 2 disks? So you copy from good boot disk to new target disk:

That's right.

Using disks all the same size, what I do is set the host with 2 disks as ZFS RAID 1. Change the pool to use IDs. Then add 2 more disks and issue the following commands (Using /dev/sdX and /dev/sdY as example for the new drives):

# Replicate existing disk (Normally /dev/sda or /dev/sdb)
sgdisk /dev/sda -R /dev/sdX
sgdisk /dev/sda -R /dev/sdY

# Randomize GUIDS
sgdisk --randomize-guids /dev/sdX
sgdisk --randomize-guids /dev/sdY

# Install GRUB
grub-install /dev/sdX
grub-install /dev/sdY

# Find the ID of the second partition(data) in /dev/disk/by-id (-part2)
ll /dev/disk/by-id/

# Add them to the pool
zpool add rpool mirror scsi-3600508b1001cea190e6d735e752-part2 scsi-3600508b1001c345f2ba190e6d735e752-part2

# Confirm everything is fine
zpool status

Now all your drives should be bootable and with the same partition scheme.

 

[mhayhurst]

no, it is correct although a more readable variant might be

sgdisk /dev/source --replicate /dev/target

to keep with the established source first, target second semantics of most UNIX commands.. getopts and thus sgdisk does not care about the order..

Yes, thank you fabian for the clarification!