Other Testing speed of USB pen drives

… Proceed with caution.

… an error (no surprise, given the low quality): …

I threw caution to the wind and gave this crummy (low quality) USB drive to OpenZFS in FreeBSD 14.0-CURRENT for a more revealing stress test.

Result:
  • explicit I/O failures – zpool-status(8) output in the xterm window
  • Konsole stopped responding when I attempted to ls(1) list the content of the mount point – there's my lack of caution.

2021-06-11 20:18:42.png




Basically I have a dodgy disk …

After restarting the operating system, my dodgy disk that was previously the sole device in the pool is no longer usable. Even lsblk has trouble, the disk (at /dev/da1) becomes non-existent:

Code:
% tail -f -n 0 /var/log/messages
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: ugen1.6: <General UDisk> at usbus1
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: umass2 on uhub4
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: umass2: <General UDisk, class 0/0, rev 2.00/1.00, addr 6> on usbus1
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: da1 at umass-sim2 bus 2 scbus7 target 0 lun 0
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: da1: <General UDisk 5.00> Removable Direct Access SCSI-2 device
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: da1: 40.000MB/s transfers
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: da1: 1911MB (3913728 512 byte sectors)
Jun 11 20:55:42 mowa219-gjp4-8570p kernel: da1: quirks=0x2<NO_6_BYTE>
^C
% time lsblk
DEVICE         MAJ:MIN SIZE TYPE                              LABEL MOUNT
ada0             0:130 466G GPT                                   - -
  ada0p1         0:132 200M efi                        gpt/efiboot0 -
  ada0p2         0:134 512K freebsd-boot               gpt/gptboot0 -
  <FREE>         -:-   492K -                                     - -
  ada0p3         0:136  16G freebsd-swap                  gpt/swap0 SWAP
  ada0p3.eli     2:57   16G freebsd-swap                          - SWAP
  ada0p4         0:138 450G freebsd-zfs                    gpt/zfs0 <ZFS>
  ada0p4.eli     0:148 450G zfs                                   - -
  <FREE>         -:-   4.0K -                                     - -
cd0              0:168  10M cd9660             iso9660/ONEPLUS%20DRIVERS -
da0              0:233  29G GPT                                   - -
  <FREE>         -:-   1.0M -                                     - -
  da0p1          0:234  29G freebsd-zfs                gpt/efiboot0 <ZFS>
stat: /dev/da1: stat: No such file or directory
stat: /dev/da1: stat: No such file or directory
da1            1.9G:-      - -                                       
0.126u 0.190s 1:21.96 0.3%      69+172k 188+0io 0pf+0w
% su -
Password:
root@mowa219-gjp4-8570p:~ # time zpool import
no pools available to import
0.000u 0.015s 0:00.23 4.3%      0+0k 9856+0io 0pf+0w
root@mowa219-gjp4-8570p:~ # man lsblk
No manual entry for lsblk
root@mowa219-gjp4-8570p:~ # time lsblk da1
NOPE: disk 'da1' does not exist in the system
…
 
I can't install anything at the moment as I have a copy of FreeBSD 11.1 installed and pkg seems to be broken. The 500GB disk is really really slow although I can't find any statistics to prove it.

I was wondering if there might be a firmware upgrade available which might speed it up. The disk itself seems to go to sleep every so often and becomes unresponsive.
 
This is one that does not work:

Code:
Vendor:               Kingston
Product:              DataTraveler 3.0
Revision:             PMAP
Compliance:           SPC-4
User Capacity:        31,004,295,168 bytes [31.0 GB]
Logical block size:   512 bytes
Physical block size:  8192 bytes
Lowest aligned LBA:   12142
Formatted with type 2 protection
128 protection information intervals per logical block
1024 bytes of protection information per logical block

It probably didn't work from new. It is supposed to be USB-3, but very rarely would do that, and write throughput is between 20 and 2000 Byte/s, anyway.
 
Re: the TOSHIBA MK5061GSYN

… The 500GB disk is really really slow although I can't find any statistics to prove it.

I was wondering if there might be a firmware upgrade available which might speed it up. …

I should not expect a firmware upgrade to make a difference in this case.
 
Here is a good article about the trouble with benchmarking and Bonnie.

bonnie++ isn't testing disk I/O – it's testing file I/O. This may correspond to disk I/O, depending on whether the I/O is read or write, cached, has synchronous flags, matches the file system record size, etc.

may not be what you would assume, given the role of bonnie++. This kind of surprise isn't uncommon with benchmarking, and is why in-flight analysis is so important.
 
… testing file I/O. …

Re: <https://forums.FreeBSD.org/threads/80629/post-516867> above, I wasn't sure how to interpret the measurements from stressdisk before I broke the drive. Stupidly I didn't take a screenshot (of measurements) before I caused Konsole to freeze. If I recall correctly:
  • completion of the first 1,000 bytes (the first test file) was reported after around four minutes
  • the measured writes dropped from around 6 MBytes per second (at the outset) to a little less, as time passed (not surprising) but not below 5 MBytes per second
  • it took more than twenty minutes to write less than three quarters of the second test file (1.57 of the 1.75 G pool allocated, 89% capacity when the device faulted in response to I/O failures).
A lazy/rough calculation: 5 x 60 x 20 = 6,000 MBytes = 5.859375 gigabyte (GB), which is much more than the capacity of a 2 GB drive :)
 
Back
Top