Solved ZFS confused by WD_BLACK SN7100 '32k blocks'

J

jb82

Hey,
I have two NVME M.2 disks, one WD_BLACK SN7100 and second WD Blue SN580. The faster black one reports 32k blocks and it gets misinterpreted by the ZFS. Please, see two diskinfo runs. Should I ignore that? No firmware updates available. I'm on 14.3-RELEASE.

Code: 
$ zpool status big
  pool: big
 state: ONLINE
status: One or more devices are configured to use a non-native block size.
    Expect reduced performance.
action: Replace affected devices with devices that support the
    configured block size, or migrate data to a properly configured
    pool.
config:

    NAME           STATE     READ WRITE CKSUM
    big            ONLINE       0     0     0
     gpt/wdblack  ONLINE       0     0     0  block size: 4096B configured, 32768B native

errors: No known data errors

Code: 
$ doas diskinfo -i /dev/nda0
/dev/nda0
    4096            # sectorsize
    2000398934016    # mediasize in bytes (1.8T)
    488378646       # mediasize in sectors
    32768           # stripesize
    0               # stripeoffset
    WD_BLACK SN7100 2TB    # Disk descr.
    2509KJ407078    # Disk ident.
    nvme0           # Attachment
    Yes             # TRIM/UNMAP support
    0               # Rotation rate in RPM

Code: 
$ doas diskinfo -i /dev/nda1
/dev/nda1
    4096            # sectorsize
    2000398934016    # mediasize in bytes (1.8T)
    488378646       # mediasize in sectors
    0               # stripesize
    0               # stripeoffset
    WD Blue SN580 2TB    # Disk descr.
    25071Y801410    # Disk ident.
    nvme1           # Attachment
    Yes             # TRIM/UNMAP support
    0               # Rotation rate in RPM
 
The interesting stuff is, that I fio-measured performance even with ashift=12 and it's just fine. I'm a bit afraid to pick 32k ashift... is that even possible? Isn't that only some kind of marketing?
 
jb82 said:
The interesting stuff is, that I fio-measured performance even with ashift=12 and it's just fine. I'm a bit afraid to pick 32k ashift... is that even possible? Isn't that only some kind of marketing?
Click to expand...
32K is a very common flash block size. Most likely this is the truth.
But it's not a very practical block size, so most devices hide it and report something more "conventional".

P.S.
With flash storage there are also multiple block sizes like erase block size, programming block size.
 
Andriy said:
32K is a very common flash block size. Most likely this is the truth.
But it's not a very practical block size, so most devices hide it and report something more "conventional".

P.S.
With flash storage there are also multiple block sizes like erase block size, programming block size.
Click to expand...
Thanks for the insights!

What would you suggest? To keep 4k and ignore the warning? What about 8k/16k would that even help in any way, there's no convergence towards 32k performance, right?
 
jb82 said:
Thanks for the insights!

What would you suggest? To keep 4k and ignore the warning? What about 8k/16k would that even help in any way, there's no convergence towards 32k performance, right?
Click to expand...
I would just keep 4K.
 
You must log in or register to reply here.
Back
Top