My replies inline.
On 10/08/2024 at 03:03, Ryan England wrote:
Is there any good documentation available for md? I'd like to be confident we're setting this up correctly.
man md
man mdadm
raid.wiki.kernel.org
On August 9, 2024 8:37:09 PM EDT, Roger Heflin <rogerheflin@xxxxxxxxx> wrote:
I may be mis-reading what a partition looks like on nvme. if nvme0n1
is the whole device, then you probably want a partition. I
NVMe naming scheme is a bit confusing. nvmeX is the physical device but
is not a block device usable to store data. IIUC nvmeXnY is some kind of
"physical partition" or "logical drive" defined inside the NVMe embedded
controller and is a block device used like any regular disk sdX. I guess
there may be several of them, but I have only seen nvmeXn1 so far.
nvmeXnYpZ is a "logical" partition (not in the sense of msdos
primary/logical partitions) like sdXY managed by usual partitioning
tools such as fdisk or parted.
How would you use parted to create a partition you would use for software RAID?
wipefs -a /dev/nvmeXn1 # remove metadata from the drive
parted /dev/nvmeXn1
create a GPT disk label and a partition with raid flag
wipefs -a /dev/nvmeXn1p1 # remove metadata from the partitions
mdadm --create ... /dev/nvmeXn1p1
The original report showed that on nvme0n1p1 there as appeared to be a
gpt partition table.
And this is what puzzles me. I do not see how it could happen, except if
someone did something extremely unwise such as running parted on
partitions /dev/nvmeXn1p1 after creating the RAID array.
You also mentioned using wipefs to wipe the metadata. Would you run the following:
- wipefs -a /dev/nvme0n1*
wipefs does not support multiple devices.
