ashift=12 enforcement for IDE (CMDK) disks

[Jim Klimov]

Hello all,

(Another +1 in favor of explicit hacky ashift or clear blocksize
setting in zpool program)

I have the zpool-12 binary stashed in my archives (the original
websites/blogs are gone), but it does not work anymore anyway:

ld.so.1: zpool-12: fatal: relocation error: file /tmp/zpool-12:
symbol zpool_set_history_str: referenced symbol not found
Killed

Googling shows that this happened back with oi151a6 already.

Unfortunately, right now I'm stuck with an IDE disk (actually a
SATA in Legacy mode) and can't use the sd.conf tricks, apparently...

Any ideas about how to lie to ZFS for an IDE/CMDK device (from
an OI LiveUSB environment)? Perhaps via loopbacks or iSCSI, but
in such a way that the on-disk raw block device would still be
bootable?

Also, just in case you know: what SSD page sizes are used by a
"Seagate Enterprise SATA SSD 120Gb (ST120FN0021)" and does it
matter much anyway for an rpool with nothing but the OS image
and maybe occasional dumping or swapping? That is, can I leave
it at the default ashift=9 and worry not about performance and
wear-out implications?

Does the ashift size matter any for ZIL and L2ARC devices?
Or, being ring buffers with bulk writes, they don't suffer much
from fragmentation and half-written sectors and pages anyway?

FWIW, I aligned and sized the partitions (rpool slice, zil and
l2arc partitions for data pool) in multiples of 32256 (63*512)
512b-sectors so that both 256Kb pages and 63-sector "tracks" are
both happy. Cylinders (63*255 = 16065) are not so happy... oh
well, I guess this only might matter for rpool's tail. There is
still time to remake and retry :)

Thanks for ideas,
//Jim

[Richard Elling]

Post by Jim Klimov
Hello all,
(Another +1 in favor of explicit hacky ashift or clear blocksize
setting in zpool program)
I have the zpool-12 binary stashed in my archives (the original
symbol zpool_set_history_str: referenced symbol not found
Killed
Googling shows that this happened back with oi151a6 already.
Unfortunately, right now I'm stuck with an IDE disk (actually a
SATA in Legacy mode) and can't use the sd.conf tricks, apparently...

IIRC, [E]IDE does not support 4k sectors. Similarly, SATA does not support
520 byte sectors. Your path forward is clear.
-- richard

Post by Jim Klimov
Any ideas about how to lie to ZFS for an IDE/CMDK device (from
an OI LiveUSB environment)? Perhaps via loopbacks or iSCSI, but
in such a way that the on-disk raw block device would still be
bootable?
Also, just in case you know: what SSD page sizes are used by a
"Seagate Enterprise SATA SSD 120Gb (ST120FN0021)" and does it
matter much anyway for an rpool with nothing but the OS image
and maybe occasional dumping or swapping? That is, can I leave
it at the default ashift=9 and worry not about performance and
wear-out implications?
Does the ashift size matter any for ZIL and L2ARC devices?
Or, being ring buffers with bulk writes, they don't suffer much
from fragmentation and half-written sectors and pages anyway?
FWIW, I aligned and sized the partitions (rpool slice, zil and
l2arc partitions for data pool) in multiples of 32256 (63*512)
512b-sectors so that both 256Kb pages and 63-sector "tracks" are
both happy. Cylinders (63*255 = 16065) are not so happy... oh
well, I guess this only might matter for rpool's tail. There is
still time to remake and retry :)
Thanks for ideas,
//Jim
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182191/22820713-4fad4b89
Modify Your Subscription: https://www.listbox.com/member/?&
Powered by Listbox: http://www.listbox.com

--

***@RichardElling.com
[phone]












-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182191/23047029-187a0c8d
Modify Your Subscription: https://www.listbox.com/member/?member_id=23047029&id_secret=23047029-2e85923f
Powered by Listbox: http://www.listbox.com

[g***@gmail.com]

This post might be inappropriate. Click to display it.

[jason matthews]

Post by g***@gmail.com
The only reason I can imagine for running in IDE mode is attempting to use seriously old system firmware or alternate operate systems on the disk (multiple OS' on a root drive). In those situations I suspect that you may be better served by limiting the contents of the root disk to the boot archive and unix kernel loadables, and then mounting the real rpool from another pool on disks that are running in SATA or SAS mode.

I have migrated my windows xp OS (circa. 2002) from pillar to post on various laptops over the years via ghost. But, since the original system was IDE, i have to run in legacy mode/IDE emulation so windows will boot...

I suspect that is not his problem -- but that is my reason for running in IDE mode :-)


j.

[Jim Klimov]

Post by jason matthews

Post by g***@gmail.com
The only reason I can imagine for running in IDE mode is attempting to use seriously old system firmware or alternate operate systems on the disk (multiple OS' on a root drive). In those situations I suspect that you may be better served by limiting the contents of the root disk to the boot archive and unix kernel loadables, and then mounting the real rpool from another pool on disks that are running in SATA or SAS mode.

I have migrated my windows xp OS (circa. 2002) from pillar to post on various laptops over the years via ghost. But, since the original system was IDE, i have to run in legacy mode/IDE emulation so windows will boot...
I suspect that is not his problem -- but that is my reason for running in IDE mode :-)

Actually, Windows 7 (unlike illumos) has no problems switching back and
forth between IDE and SATA modes of the boot device. Earlier Windows
might need the driver preinstalled, and could change mode as well then.

In my case, this is a "SATA/ODD" port on the HP N54L where my brother is
reluctant to change the BIOS so that proper SATA is supported on this
port so as to not void the warranty (waiting for delivery of a proper
HBA anyway), but we wanted to start out with installation of the OS.
AFAIK, we're lucky that the drive is accessible on this port at all ;)
Changing the boot mode (device path) later would then be a matter of
booting the Live media, import, export, reboot.

Post by jason matthews
IIRC, [E]IDE does not support 4k sectors. Similarly, SATA does not support
520 byte sectors. Your path forward is clear.
-- richard

I am not sure if ashift=12 automatically means 4k sectors, does it?
It means (AFAIK) the minimal transaction size and consequently block
alignment (after all, it can be used on 512b/512e HDDs or LUNs).

//Jim

[Volker A. Brandt]

Hi Jim!

Post by Jim Klimov
In my case, this is a "SATA/ODD" port on the HP N54L where my brother is
reluctant to change the BIOS so that proper SATA is supported on this
port so as to not void the warranty

You could always re-flash. Of course if the box is broken that's
not an option. :-) However, except for the enabled extra options
in the BIOS there is no change; the BIOS will still identify itself
by the original HP version. The HP service might not even notice
any difference.

We have immediately flashed all boxes after unpacking them; indeed
without the availability of the BIOS mod we would not have bought
the microservers.

As others have said, these boxes are disposable, as it is cheaper to
buy a new one than having any failed component repaired or replaced
by HP.


Regards -- Volker

[Freddie Cash]

Post by g***@gmail.com

Post by g***@gmail.com
The only reason I can imagine for running in IDE mode is attempting to

use seriously old system firmware or alternate operate systems on the disk
(multiple OS' on a root drive). In those situations I suspect that you may
be better served by limiting the contents of the root disk to the boot
archive and unix kernel loadables, and then mounting the real rpool from
another pool on disks that are running in SATA or SAS mode.
I have migrated my windows xp OS (circa. 2002) from pillar to post on
various laptops over the years via ghost. But, since the original system
was IDE, i have to run in legacy mode/IDE emulation so windows will boot...
I suspect that is not his problem -- but that is my reason for running in IDE mode :-)

Since you have the ghost image as a backup, the following will fix your
issues:
- boot into safe mode
- go into device manager
- remove all the disk controllers
- reboot into bios
- set disks to SATA/AHCI/whatever
- boot into Windows and let it install the correct drivers

Migrated a few systems from IDE to SATA that way.​​
--
Freddie Cash
***@gmail.com



-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182191/23047029-187a0c8d
Modify Your Subscription: https://www.listbox.com/member/?member_id=23047029&id_secret=23047029-2e85923f
Powered by Listbox: http://www.listbox.com

[jason matthews]

Post by Freddie Cash
- boot into safe mode
- go into device manager
- remove all the disk controllers
- reboot into bios
- set disks to SATA/AHCI/whatever
- boot into Windows and let it install the correct drivers
Migrated a few systems from IDE to SATA that way.​​

i will give that a whirl tonight. that sounds like solid wisdom. i made a previous effort to get xp to use sata but who knows, maybe i wasn't at the required patch level for it to succeed.


just to tie this into illumos -- all the windows computers in the house back up with ghost, over iscsi to the file server. there a virtual box instance that fires up nightly with clones of ghost data and exports the ntfs filesystem over cifs for recovery purposes.

okay, my conscious is clear -- it all ties together -- even if it exposes my pathological thought process.

thanks for the tip.

j.





-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182191/23047029-187a0c8d
Modify Your Subscription: https://www.listbox.com/member/?member_id=23047029&id_secret=23047029-2e85923f
Powered by Listbox: http://www.listbox.com

[Richard Yao]

Post by jason matthews

Post by Freddie Cash
- boot into safe mode
- go into device manager
- remove all the disk controllers
- reboot into bios
- set disks to SATA/AHCI/whatever
- boot into Windows and let it install the correct drivers
Migrated a few systems from IDE to SATA that way.​​

i will give that a whirl tonight. that sounds like solid wisdom. i made a previous effort to get xp to use sata but who knows, maybe i wasn't at the required patch level for it to succeed.

I have done this myself and it worked for me.

Post by jason matthews
just to tie this into illumos -- all the windows computers in the house back up with ghost, over iscsi to the file server. there a virtual box instance that fires up nightly with clones of ghost data and exports the ntfs filesystem over cifs for recovery purposes.

It is possible to boot Windows off iSCSI. It is harder to do it with Windows XP than it is to do with Windows 7. Windows 7 comes with a built-in iSCSI initiator. You just need to disable a network filter that prevents the iSCSI information from being passed to it, transfer the image to an iSCSI LUN and boot off it.

Post by jason matthews
okay, my conscious is clear -- it all ties together -- even if it exposes my pathological thought process.
thanks for the tip.
j.
illumos-zfs | Archives | Modify Your Subscription

-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182191/23047029-187a0c8d
Modify Your Subscription: https://www.listbox.com/member/?member_id=23047029&id_secret=23047029-2e85923f
Powered by Listbox: http://www.listbox.com