Problem with WD Advanced Format drive in LINUX (WD15EARS)

[ramiro_morales]

---

klubad wrote:

I bought an external WD HD yesterday, a My Book Essential 1.5T. The HD in the box is a WDC WDC WD15EARS-00Z5B1

hdparm -I /dev/sdb

/dev/sdb:

ATA device, with non-removable media
    Model Number:       WDC WD15EARS-00Z5B1                     
    Serial Number:      WD-WMAVUxxxxxxx
    Firmware Revision:  80.00A80
[...]
    Logical  Sector size:                   512 bytes
    Physical Sector size:                  4096 bytes
    Logical Sector-0 offset:                  0 bytes 

---

Wait, there seem to be something different (and hopefully good) in what klubat has posted.

It seems the 1.5 TB drive he got is correctly reporting the 4kB physical block size. This is something I haven't seen reported by any other Advance Format EARS drive owner in this thread or other Internet discussions so far.

Could it be that WD has changed things in the "00Z5B1" models to correctly report the physical block size for the benefit of the Linux-based partitioning tools and so help people avoid getting non 4kB-aligned partitions?. Or maybe the difference is in the fact he's is the drive inside a external My Book Essential?.

For comparison, I've just bought an internal 1 TB WD10EARS 64MB cache Caviar Green drive with the same firmware version as his (80.00A80) but with a 00Y5B1 suffix in the model number that, just like all other posters' (here and other Internet discussions), reports  512B both for the physical and logical blocks sizes:

$ sudo ./hdparm -V
hdparm v9.28

$ sudo ./hdparm -I /dev/sdb

/dev/sdb:

ATA device, with non-removable media
        Model Number:       WDC WD10EARS-00Y5B1
        Serial Number:      WD-WMAV5*******
        Firmware Revision:  80.00A80
[...]
Configuration:
[...]
        Logical/Physical Sector size:           512 bytes

My kernel version is 2.6.33.x. and the drive has no jumper between pins 7-8.

[Bill_S]

It's emulating the 512 bytes, it's really 4k. 

[Miles_Prower]

Hello,

I'm trying to format my new WD15EARS but I'm having some trouble with it. Could someone help me out on this issue?

This is what I'd like to obtain :

(I hope my calculations are correct)

Since I'd like to have more than 4 partitions on this HDD for pratical purposes, I have to use Extended partitions (noted (E) above) to cram two logical drives inside. I can't figure out how to do this in Parted, and other tools did not allow me to align properly the partitions.

I tried to use GParted to set up the partitions as I wanted, all in Fat32, then tried to use WDAlign to align the partitions and expected to change their format after that. The WD Align tool did not work, the first partition started at 63 (which is bad).

Any help would be appreciated !

Thanks

[frostschutz]

---

ramiro_morales wrote:

Could it be that WD has changed things in the "00Z5B1" models to correctly report the physical block size for the benefit of the Linux-based partitioning tools and so help people avoid getting non 4kB-aligned partitions?. Or maybe the difference is in the fact he's is the drive inside a external My Book Essential?.

I just got a WDC WD15EARS-00Z5B1 (internal drive) and it reports 512 bytes as logical / physical sector size. Using kernel 2.6.34-rc6 and an Intel DP35DP board. So nothing changed.

Getting the alignment correctly on this one was quite a challenge.

GPT partition table (32 partitions), parted couldn't get the align right by itself, so I wrote a small Python script that calculated partition sizes and alignment for me. (Can be done manually if it's 2-3 partitions but gets tedious for a larger number).

cryptsetup/LUKS: is luckily automatically aligned to 4k boundaries

LVM: is luckily automatically aligned to 4k boundaries

XFS filesystem: would know to use 4K blocks if the drive reported it correctly, which it doesn't, so you have to supply the sector size as option to mkfs.xfs. Especially the XFS problem I noticed too late and as a result I spent another day shuffling data around, remaking all my XFS volumes.

However now that all things are done and my encrypted xfs lvm volumes are now running properly aligned on this drive, I'm very happy with its performance. I hope the next generation of drives will report their sector sizes correctly!

[caspercba]

Hi all,

Using the information provided by hans 1967 I could format my hd and use it on my linux box.

My disc: WD10EARS ( 1tb).

My problem: I/O operations blocking my machine for the most stupid tasks. This behaviour was exponential woring with svn trunks and many little files. Doing a svn cleanup command would take many minutes to finish.

Error logs: None. I could not even find one error in sata controllers or IO read/write errors, so I pointed my efforts on HD formatting.

HD format was initially done using cfdisk and creating parttions in any place and size. Even using 4k blocks ( by default is in 4k), this was not working ok.

Solution:

1. run fdisk with the following params:

#fdisk -H 224 -S 56 /dev/sda

2. Create partitions.

- NOTE: Use starting cylinders multiples of 4.

3. The first partition is usually located at cylinder 63 ( leaving some free mb at disk start ).

    As the start cylinder has to be multiple of 4, I had to change it. I preferred to use 64 as start cylinder, didn't have so much time to test, but 20 would have worked fine. I spent 1 week installing gentoo with this slow disk and reinstalled after the fix

    SO: Start cylinder 64.

4. For partition end cylinder I preferred to specify size and not cylinder using the format +[value]G , of course replace [value] by the size you want.

5. Done. this is my current table ( notice I dont have swap partition hehe, Im just to create one):

#fdisk -l /dev/sda

Disk /dev/sda: 1000.2 GB, 1000204886016 bytes
224 heads, 56 sectors/track, 155733 cylinders
Units = cylinders of 12544 * 512 = 6422528 bytes
Disk identifier: 0x000ea1b2

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *          64         146      520576   83  Linux       --> notice start cylinder is 64 ( multiple of 4)
/dev/sda2             148       16866   104861568   83  Linux     --> notice start cylinder is  multiple of 4
/dev/sda3           16868       75382   367006080   83  Linux   --> notice start cylinder is multiple of 4
/dev/sda4           75384      155733   503955200    5  Extended   --> start cyl is multiple of 4 and covers the rest of the disk
/dev/sda5           75384       92102   104861540   83  Linux   --> start cyl is multiple of 4 and starts in the same place as the extended partition since its a logical unit.

NOTE: Again, the ending cylinder was chosen by fdisk, since I only specified starting cylinder and size.

Good luck to all the linux community. This problem was really annoying for me and I spent too much hours before discovering the disk was ok, and the kernel modules were also ok.

Thanks to all this forum. Thanks to  hans1967  specially, he gave the easiest directives to solve this.

My worst regards to the WD workers who wrote in the harddrive cover:

"For windows installations please align the disk using the utillity. For linux users: IS READY TO USE AS IS" WTF??

Regards,

Gaspar

[sysmg]

I just purchased an WD advanced format 2TB drive (WD20EARS). I am using it with Debian 4 on an old Pentium 2 computer. It is really slow. I set it up using fdisk -u and starting the partition at 64 and then using MKFS to format an ext3 filesystem specifying a blocksize of 4096. Unfortunately that did not avoid the performance issue. Does anybody have any ideas? Should I try resetting the jumpers to turn off 4k blocksize? I am trying to decide if I should RMA the drive.

Here is how I can tell the drives are slow. The system has an IDE system drive 12GB w/about 6GB free. It also has a 1TB SATA WD internal drive (WD10EACS). I have added the new 2TB SATA WD (WD20EARS) advanced format drive externally via eSATA. I intend to replace the 1TB drive with the 2TB driver. I copied the data from the 89% full 1TB drive to the new 2TB drive. It took days at under 4MB/s. So I decided to compare the performance of the 1TB and the 2TB. I copied a 5GB file from the 1TB to the IDE drive which averaged about 8MB/s and copying the same file from the 2TB to the IDE drive averaged only 2.9MB/s.  Tried copying the file from the IDE drive to the 1TB at about 9MB/s, and from the IDE drive to the 2TB at about 2.9MB/s.

-Stew

[sysmg]

Wow! I upgraded to Debian 5.03 and a Pentium 4 processor and now copying the same file from 2TB EARS to 80GB internal IDE drive is at 83MB/sec and copying from 80 GB IDE to 2TB is 46MB/sec.  Now I can leave it with 4k sectors.

-Stew

[leviathan]

I prepare to buy WD15EARS (with 2,930,277,168 sectors) and I don't understand why in these operations

(parted) mkpart
Partition name?  []?
File system type?  [ext2]?
Start? 64
End? -1

argument "-1" is at "end". Because 2,930,277,167th is a logical last (8'th of 8) block of a one physical sector, isn't it? But why system tells about 2930277134th sector:

Warning: You requested a partition from 64s to 2930277167s.
The closest location we can manage is 64s to 2930277134s.

?

It doesn't a last, just a 7th of 8.

[oelauge]

I tried all steps (described here) with parted and fdisk to configure my WD20EARS under Linux with the actual kernel.

My results are to every time the same. I see 512 bytes as phyiscal and logical and hdparm shows me a value with 50 MB/s. I think this is too slow.

What should i do?

[Alfredo]

Hi all,

I did use Gparted-LiveCD test version 0.6.2-2.

See:

http://ziggo-gebruikers.nl/forum/showthread.php?t=16533&page=56

Nb!.  Be sure, you have to "resize" the first partition by 1 Mb; this might be a little "bug" in Gparted.

But it looks like you can partition and format your WD20EARS correctly in Advanced Format in Gparted now !

Aligned to Mib. ?

With "fdisk -lu /dev/sdd" I did see:

Disk /dev/sdd: 2000.4 Gb, 2000398934016 bytes
255 Heads, 63 sectors/tracks, 243201 cylinders, total 3907029168 sectors
units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes/512 bytes
Disk identifier: 0x0002decd

Device Boot Start End Blocks Id System

/dev/sdd1 2048 4196351 2097152 83 Linux
Partition 1 does not end on cylinder boundary.  *
/dev/sdd2 4196352 4405247 104448 83 Linux
Partition 2 does not end on cylinder boundary.  *
/dev/sdd3 4405248 38454600191 1925097472 83 Linux
/dev/sdd4 38454600192 3907028991 26214400 83 Linux

The * remarks by "Fdisk" are not correct, I do think!

With "sfdisk -uB -l /dev/sdd":

Disk /dev/sdd : 243201 cylinders, 255 heads, 63 sectors/track

units = blocks of 1024 bytes, counting from 0

Device Boot Start End #blocks Id System

/dev/sdd1 1024 2098175 2097152 83 Linux

/dev/sdd2 2098176 2202623 104448 83 Linux

/dev/sdd3 2202624 1927300095 1925097472 83 Linux

/dev/sdd4 1927300096 1953514495 26214400 83 Linux

I did re-format the disk with "mkfs.ext3 -b 4096 /dev/sddx" to be sure that 4K blocks where used.

Partition 1, 3 and 4 automatic did format with 4K blocks with "mkfs.ext3 /dev/sddx"; only partition 2 required the "-b 4096" parameter.

But this all did not make a difference for the WD Align Utility; and "sfdisk", they did both present the same information - before and after.

With the "WD Align Utility - by Acronis" I could see in Windows XP that the disk was Optimally Aligned # ! (The utility displays this information)

The result:

                                                                           Cap              Used                         Start                            End

Disk 4

#  Ext3 (Unlabeled) (H: )             Pri             2 Gb              99,87 Mb Ext3           2048                           4.196.351                \Device\000000921

#  Ext3 (Unlabeled) (J: )              Pri         102 Mb             9,392 Mb Ext3           4.196.352                  4.405.247                 \Device\000000922

#  Ext3 (Unlabeled) (L: )              Pri      1,793 Tb             29,00 Gb Ext3           4.405.248                  3.854.600.191         \Device\000000923

#  Ext3 (Unlabeled) (S: )             Pri            25 Gb             574,4 Mb Ext3           3.854.600.192          3.907.028.991         \Device\000000924

Also the "WD Align System Utility - by Paragon", tells me, the disk is optimally aligned.

So I think Gparted LiveCD 0.6.2-2 can do the job, my disk "looks" in Advanced Format correctly !

But I'm in doubt, is my HD realy in 4 Kb sector or cluster size?

Last info: I did prepare this disk for replacing my HD 320Gb in my PVR Humax iHDR 5050-c.

And this works. But I like the challenge to use this WD20EARS correctly in AF and aligned.

What is your opinion?

Alfredo

The Netherlands