Kernel lost device (good disk)

[def]

Hi all,

I'm having some trouble with my new NAS build.
Hardware:

- Intel celeron G550
- Asrock B75 pro3-M
- Intel gigabit CT adapter
- Kingston KVR1333D3N9/8GB (1 stick)
- Be quiet L7 300W power supply
- 2x WD15EARS in a ZFS mirror. (on the onboard sata II ports: from intel chipset)


The problem is that one of the disks gets losts. But after a reboot it's back online.
I checked the disk with the WD tool (long check). Used it with Windows = perfect.
I had the same problem with freenas 8.3, and hoped a upgrade to N4F would help, but it didn't.

log from 4 december:

(happend 10 times)

please help me solve this problem

Thanks,
Wouter

[raulfg3]

Check in BIOS how SATA is configured, AHCI is the recomended way.

Do not use external SATA enclosure, connect disk to motherboard using a good SATA cable with metalic clips, and as sort as posible.

[def]

Hi thanks for the reply.

I forgot to mention that the bios was set on AHCI, (before install ofc).
I used 2 new SATA(III) cables from another Gigabyte MB. Disks are just installed in the case.

[al562]

Hi Def,

Welcome to NAS4Free forum. Thank you for the information provided. Since this is your first topic, I will still remind you to please read and follow the Forum Rules & Guidelines and the FAQs.

Raul is correct, you need to focus on the BIOS settings of this MB or the onboard SATA controller settings. Try telling the BIOS there are no drives installed, let NAS4Free find them on its own, with some MB's this works best. According to the Manufacturer this MB has 2 different controllers with different features:

• 1 x SATA3 6.0 Gb/s connector by Intel® B75, supports NCQ, AHCI and Hot Plug functions
• 2 x SATA3 6.0 Gb/s connectors by ASMedia ASM1061, support NCQ, AHCI and Hot Plug functions

and 8 ports:

• 5 x SATA2 3.0 Gb/s connectors, support NCQ, AHCI and Hot Plug functions
• 3 x SATA3 6.0 Gb/s connectors

You'll notice the manufacturers' description of the 6.0 GB/s ports are contradictory. With this combination of different features you need to expect the BIOS/controllers will automatically change the order of reported devices as you plug new ones in. There is little that can be done to predict the results, unless the manufacturer gives you all the details. You will need to experiment and become familiar with the MB/features, especially if there is no way to have the BIOS and Controller give direct access to NAS4Free.

Let us know what you find,
Al

[raulfg3]

Use Intel b75 SATA only ( black connectors) do not mix asmedia & Intel controller until You are totally sure, where are the problem.

[def]

Hi sorry, I forgot some info.
(The N4F FAQ is down atm, gives an internal 500 error)

The disks are on the Intel B75 chipset. (SATA 1 and SATA 2)
SATA 0 (the first gray one) is the only SATA III from the intel chipset and is free for a cache SSD.

The ASmedia SATA III ports are at another place on the mobo and are not in use.

I only have 2 disks. (In the future I would add 3x3(4)TB for media, but then this should work fine first ) N4F is installed on a USB flash drive (embedded).

N4F info:
Version: 9.1.0.1 (x64)
Revision: 531

BIOS (UEFI) settings:
CPU config:
- C6 state: enabled
- C3 state: enabled
- C1E state: enabled
- package C state support: Auto
- Hardware prefetcher: Enabled
- Adjacent Cache Line Prefetch: Enabled
- Thermal Throtlling: enabled
- No-execute memory protection, and intel virtualization technology are disabled.

Storage configuration in Bios:
- SATA controllers: enabled
- SATA Mode Selection: AHCI
- SATA Aggressive Link Power Management: disabled
- Hard Disk SMART: Enabled
- ASmedia SATA ports (just set them disabled. we'll see, but I don't think this helps)

On the intel ports I can change: "external SATA" which is disabled on both ports. There is an option SATA device Type (Hard disk or SSD). Maby I should set this on SSD? (is HDD now). (AF drives: I'll maby try this later on). I don't see an option where I could let N4F detect the drivers.

Some relevant info from the N4F report:

Code: Select all

ATA disk:
---------
atacontrol:
[b]ATA_CAM option is enabled in kernel.
Please use camcontrol instead.[/b] 
[i]there are some CAM errors in the log, something todo with this? easy way to change this on an embedded install?[/i]

SCSI disk:
----------
<WDC WD15EARS-00J2GB0 80.00A80>    at scbus3 target 0 lun 0 (ada0,pass0)
<WDC WD15EARS-00MVWB0 51.0AB51>    at scbus4 target 0 lun 0 (ada1,pass1)
<pny USB 2.0 FD 1.00>              at scbus9 target 0 lun 0 (pass2,da0)

Thanks again,
Wouter

[al562]

Hi Wouter,

there are some CAM errors in the log, something todo with this? easy way to change this on an embedded install?

If you think all BIOS settings are correct and you continue to get these errors (like in your screenshot in first post here) I would check the cables and connectors. Maybe bad controller? I have not seen that particular error before, but I would not try to turn off CAM Control, the disk should not disappear like that and if it is not caused by settings it is most likely a hardware problem .

Regards,
Al

[def]

Hi,

I think all bios settings are OK. Then the hardest thing comes What controller is bad? ( I thought of that before)
The controller on the hard disk (my first thought) or the chipset sata controller on the MOBO?
I conntacted WD but they didn't support Linux. (uhum UNIX based systems). And I can't send a good drive RMA :s
I also tried getting another WD15EARS drive from the local store to see if the disk was the problem but he didn't have one.
My old hardware i used for a freenas 0.7.X build is allready gone, so i can't try that mobo.

thanks again,
Wouter

[al562]

Try switching the disks around, see if the error follows a disk or if it stays with the same address.

[def]

Haha good idea, why i didn't think of that

just done + import and clear. I'll let you guys know what the result is.

[def]

Today the disk got lost again. The same as always but on another port. So i guess it's the disk. I'll contact WD and hope they don't play hard

[al562]

Try running WD's diagnostic software on that disk while still in your server. Do the long test and see if reports any errors, WD will probably have you do this anyway before they offer to replace the drive. Any error you get may be helpful, but if you get no errors then it's back to more testing .

Regards,
Al

[def]

Update:
- Ran WD diagnostic tool from bootable CD, when the disk was in the server.
- Ran 2 times (extended test), first time I got a error O223 (WD=errors found and repaired, don't know what kind of errors. Maby bad sectors, but it was clear before with HD tune check). After this i got 3 days of uptime till yesterday :s
- After having contact with WD, the drive goes RMA. (if the shipment isn't more then the drive is worth )

I'll post another update when it's solved or not

[al562]

Well, that certainly seems to indicate the disk is at fault. Just to be sure I would run the test on a completely different computer, this eliminates everything except the disk and lets you know your server is 100% OK.
I just had a WD black, 1Tb drive die. It could still be written to, but had trouble reading (errors: Read Element Failure - Too many errors to repair). Wasn't even in use for 8 months, slowed the server down to a crawl and caused it to crash. The warranty is good until the middle of 2015 but the trouble it caused is not worth it. There is no meaningful quality control at most major companies these days so you have to be very careful with the hardware.

Regards,
Al

[def]

Got a new drive, but I don't get it in the old ZFS mirror like it should.



I get the new drive in my old zfs mirror with zfs replace, but not as a .nop device ("4k sector trick")

I tried:
gnop create -S 4k ada1
zpool replace zfsmirror 1175.. ada1.nop
cannot open 'ada1.nop': no such GEOM provider


In the webGUI the file format is "Unknown or unformatted ", and I can't format the drive as zfs storage device (not displayed).
I can't add the device to:
Disks->ZFS->Pools->Virtual device->Edit


I guess i fixed it


Old drive was offline,so I did a detach of it. Then did a attach of the newDrive.nop.
(zpool attach zfsmirror WorkingDrive.nop NewDrive.nop )


I'll edit the post again if it the real problem is fixed (or not).

[al562]

Looks like everything is good, glad you got it fixed.
Al

[fsbruva]

def-

I can't tell from the specs online - do your drives actually have 4K sectors?

Code: Select all

smartctl -i /dev/ada1

Looking for: Sector Sizes: 512 bytes logical, 4096 bytes physical.

If so, what is the alignment of your pool?

Code: Select all

zdb | grep ashift

Looking for: ashift: 12

If your alignment shift is 12 and you have 4096 byte sectors, you can make a substantial improvement to disk performance by following my post: viewtopic.php?f=59&t=1494

[def]

hi, thanks for the reply.

The WD EARS serie are indeed 4kB disks (the first ones I think, and not the best series ^^).
And when i'm wright, they still report as 512 sector disks. (diskinfo -i /dev/ada0 --> Sector Size: 512 bytes logical/physical)
When i created the pool it created a .nop:

Code: Select all

diskinfo -v ada0
ada0
        512             # sectorsize
        1500301910016   # mediasize in bytes (1.4T)
        2930277168      # mediasize in sectors
        4096            # stripesize
        0               # stripeoffset
        2907021         # Cylinders according to firmware.
        16              # Heads according to firmware.
        63              # Sectors according to firmware.


diskinfo -v ada0.nop
ada0.nop
        4096            # sectorsize
        1500301910016   # mediasize in bytes (1.4T)
        366284646       # mediasize in sectors
        0               # stripesize
        0               # stripeoffset
        363377          # Cylinders according to firmware.
        16              # Heads according to firmware.
        63              # Sectors according to firmware.

The ashift is also set to 12.

Code: Select all

 zdb | grep ashift
            ashift: 12

So I hope that did the trick

[fsbruva]

Wait, your comments don't make any sense to me...
"diskinfo -i /dev/ada0 --> Sector Size: 512 bytes logical/physical"
But then you say they are 4K disks......... ?

[def]

Yes it are 4kB disks.
Tom's hardware about WD EARS series:
Samsung has had a few hard drives that use sector sizes larger than the usual 512 bytes, such as the Spinpoint M7E and the portable S1 Mini. However, it is Western Digital that has launched a new product line, the EARS-series, and moved 4KB sectors into the mainstream.
But whem I'm right these drives lie about the sector size, see this topic on WD community:
http://community.wdc.com/t5/ideas/v1/id ... 1#comments