Re: Aborted commands with arcmsr and 2xWD1500ADFD in RAID1

[Aron Stansvik]

Hello again.

2008/3/14, Aron Stansvik <elvstone@xxxxxxxxx>:
> 2008/2/26, nickcheng <nick.cheng@xxxxxxxxxxxx>:
>  > Hi Aron,
>
> >  Thanks for your patience.
>  >  If you still got into trouble, please let me know.
>  >  Thank you again,
>
>
> I have now tried:
>
>  * Turning on/off NCQ in the Areca RAID.
>  * Turning on/off read-ahead cache in the Areca RAID.
>  * Putting the disks in anti-vibration mounts in 5.25" slots.
>  * Switching SATA cables.
>  * Using legacy ATA power connectors instead of the SATA ones.
>
>  But I still have the problem. The power supply is 650W so there should
>  be plenty of power. There's only two Raptor disks, an Opteron CPU and
>  an nVidia 6600GT in the machine.
>
>  The Raptor two Raptor disks have different firmware on them, could
>  that cause any problem?
>
>  Two people who had read my post here on LKML have contacted me on
>  e-mail and have the same problem, but they have Seagate and Samsung
>  disks, and use the 1220 controller.
>
>  The problem is hard to trigger, I've not been able to trigger it with
>  any benchmarking tool, but in ~95% of the cases I can trigger it by
>  just copying a directory with lots of small files (around 500 MB).
>
>  Anyone else seeing this? I'd really like to get it to work since this
>  is my only computer :(
>
>  Should I try with XFS or ReiserFS instead of EXT3?

I've now tried with ReiserFS, same problem. My distribution didn't
have XFS as a choice at installation so I haven't tried it yet. I'll
try with a LiveCD that supports XFS later. The two disks in the array
are the only ones in the system.

Aron

>
>  Regards,
>
> Aron
>
>
>  >
>  > -----Original Message-----
>  >  From: Aron Stansvik [mailto:elvstone@xxxxxxxxx]
>  >
>  > Sent: Tuesday, February 26, 2008 6:52 AM
>  >  To: erich
>  >  Cc: nick.cheng@xxxxxxxxxxxx; akpm@xxxxxxxxxxxxxxxxxxxx;
>  >  linux-scsi@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
>  >  Subject: Re: Aborted commands with arcmsr and 2xWD1500ADFD in RAID1
>  >
>  >  Hi Erich.
>  >
>  >  2008/2/25, nickcheng <nick.cheng@xxxxxxxxxxxx>:
>  >  > Hi Aron,
>  >  >  From our field experiences and customers' feedbacks, all of them direct
>  >  to
>  >  >  vibration and power issues.
>  >  >  The vibration could be caused by FANs not only by themselves.
>  >
>  >  Okay. I have a chassi fan that is quite close to the drives, I will
>  >  try disabling it. I've also ordered two Nexus TwinDisk anti vibration
>  >  harddrive mounts with which I'll place the disks in my 5.25" slots
>  >  instead, away from any fans.
>  >
>  >  If this doesn't work, I'm stumped, as I really don't think it's the
>  >  power supply and I don't have the money to buy a new one.
>  >
>  >  >  You mentioned it could be the F/W issue.
>  >  >  If the environment does not meet the prerequisite, FW could not work
>  >  >  correctly.
>  >  >  Actually FW just reacts to the situations not it causes the issue.
>  >
>  >  Of course, I understand this. Just trying to figure this problem out..
>  >
>  >  >  Please check it out!!
>  >
>  >  I'll report back with my findings with moving disk away from fans and
>  >  using anti-vibrations mounts.
>  >
>  >  Thanks for taking your time to reply.
>  >
>  >  Aron
>  >
>  >  >  Thank you,
>  >  >
>  >  >
>  >  >  -----Original Message-----
>  >  >  From: Aron Stansvik [mailto:elvstone@xxxxxxxxx]
>  >  >  Sent: Sunday, February 24, 2008 1:54 AM
>  >  >  To: nick.cheng@xxxxxxxxxxxx
>  >  >  Cc: erich; akpm@xxxxxxxxxxxxxxxxxxxx; linux-scsi@xxxxxxxxxxxxxxx;
>  >  >  linux-kernel@xxxxxxxxxxxxxxx
>  >  >
>  >  > Subject: Re: Aborted commands with arcmsr and 2xWD1500ADFD in RAID1
>  >  >
>  >  >  Hello again Areca and LKML hackers.
>  >  >
>  >  >  2008/2/18, Aron Stansvik <elvstone@xxxxxxxxx>:
>  >  >  > Hello Nick.
>  >  >  >
>  >  >  >  Sorry that I'm not answering until now. I've been busy.
>  >  >  >
>  >  >  >  2008/2/13, nickcheng <nick.cheng@xxxxxxxxxxxx>:
>  >  >  >
>  >  >  > > Hi Aron,
>  >  >  >  >  From our experience and some customers' feedback, your issue could
>  >  be
>  >  >  caused
>  >  >  >  >  by power instability or vibration to your HDs.
>  >  >  >  >  Please check step by step:
>  >  >  >  >  (1).under your original environment, increase the SCSI command
>  >  value,
>  >  >  >  >  default=30, with the shell script, set_scsicmd_timeout(). 90 or 120
>  >  is
>  >  >  >  >  enough.
>  >  >  >  >  (2).if method 1 does not work, find out the vibration source or
>  >  change
>  >  >  the
>  >  >  >  >  power supply
>  >  >  >
>  >  >  >
>  >  >  > I will try to increase that value. I don't think it's vibration; the
>  >  >  >  disks are firmly in place in a very heavy chassi (Silverstone
>  >  >  >  SST-TJ05B-T). And I really don't think there's something wrong with
>  >  >  >  the power supply, it's a pretty new Silverstone ST65ZF 650W. This is
>  >  >  >  my own personal workstation, so I don't just have another power supply
>  >  >  >  to test with :/
>  >  >  >
>  >  >  >  I will report back on my success/failure. Thanks for your answer.
>  >  >
>  >  >  I've now tried with both 90 and 120 for the timeout value, and the
>  >  >  problem still persists. It seems to happen when lots of small writes
>  >  >  are occuring, e.g. when installing something.
>  >  >
>  >  >  I really don't think the disks are vibrating, I don't see how they
>  >  >  could. One more thing I'm going to test is to use the legacy ATA power
>  >  >  connector instead of the SATA power connector. This was what I was
>  >  >  using before when I only had a single drive and no RAID controller.
>  >  >  Maybe my power supply is malfunctioning and not giving enough power on
>  >  >  the SATA power connectors.. but I doubt it.
>  >  >
>  >  >  Is there anything else that could cause this? Have you guys at Areca
>  >  >  tested the ARC-1200 with Raptors in RAID1?
>  >  >
>  >  >  :(
>  >  >
>  >  >  Regards,
>  >  >  Aron
>  >  >
>  >  >  >
>  >  >  >
>  >  >  >  Aron
>  >  >  >
>  >  >  >
>  >  >  >  >  If your still have any questions, please feel free to let me know.
>  >  >  >  >
>  >  >  >  >  P.S. The attached driver source, arcmsr-1.20.00.15-71224, has been
>  >  >  >  >  upstreamed to kernel.org and will be released in kernel 2.6.25. If
>  >  you
>  >  >  like,
>  >  >  >  >  you could update your driver with it.
>  >  >  >  >  It fixes some minor bugs, but these bugs are nothing to do with
>  >  your
>  >  >  issue.
>  >  >  >  >
>  >  >  >  >
>  >  >  >  >  -----Original Message-----
>  >  >  >  >  From: erich [mailto:erich@xxxxxxxxxxxx]
>  >  >  >  >  Sent: Wednesday, February 13, 2008 4:33 PM
>  >  >  >  >  To: (????)???
>  >  >  >  >  Subject: Fw: Aborted commands with arcmsr and 2xWD1500ADFD in RAID1
>  >  >  >  >
>  >  >  >  >
>  >  >  >  >
>  >  >  >  >  ----- Original Message -----
>  >  >  >  >  From: "Andrew Morton" <akpm@xxxxxxxxxxxxxxxxxxxx>
>  >  >  >  >  To: "Aron Stansvik" <elvstone@xxxxxxxxx>
>  >  >  >  >  Cc: <linux-kernel@xxxxxxxxxxxxxxx>; <linux-scsi@xxxxxxxxxxxxxxx>;
>  >  >  "erich"
>  >  >  >  >  <erich@xxxxxxxxxxxx>
>  >  >  >  >  Sent: Wednesday, February 13, 2008 4:03 PM
>  >  >  >  >  Subject: Re: Aborted commands with arcmsr and 2xWD1500ADFD in RAID1
>  >  >  >  >
>  >  >  >  >
>  >  >  >  >  >
>  >  >  >  >  > (cc's added)
>  >  >  >  >  >
>  >  >  >  >  > On Mon, 11 Feb 2008 17:44:08 +0100 "Aron Stansvik"
>  >  >  <elvstone@xxxxxxxxx>
>  >  >  >  >  > wrote:
>  >  >  >  >  >
>  >  >  >  >  >> Hello LKML.
>  >  >  >  >  >>
>  >  >  >  >  >> Under semi-high disk I/O (e.g. installing a compiled KDE), I get
>  >  >  the
>  >  >  >  >  >> following (accompanied by seconds of lock-ups on the machine):
>  >  >  >  >  >>
>  >  >  >  >  >> [ 7727.345183] arcmsr0: abort device command of scsi id = 0 lun
>  >  = 0
>  >  >  >  >  >> [ 7730.348776] arcmsr0:                 scsi id = 0 lun = 0 ccb
>  >  =
>  >  >  >  >  >> '0xdfb461c0' poll command abort successfully
>  >  >  >  >  >> [ 8053.795943] arcmsr0: abort device command of scsi id = 0 lun
>  >  = 0
>  >  >  >  >  >> [ 8056.799528] arcmsr0:                 scsi id = 0 lun = 0 ccb
>  >  =
>  >  >  >  >  >> '0xdfb595e0' poll command abort successfully
>  >  >  >  >  >> [ 8884.592810] arcmsr0: abort device command of scsi id = 0 lun
>  >  = 0
>  >  >  >  >  >> [ 8887.596392] arcmsr0:                 scsi id = 0 lun = 0 ccb
>  >  =
>  >  >  >  >  >> '0xdfb56d80' poll command abort successfully
>  >  >  >  >  >> [ 8917.760216] arcmsr0: abort device command of scsi id = 0 lun
>  >  = 0
>  >  >  >  >  >> [ 8920.763797] arcmsr0:                 scsi id = 0 lun = 0 ccb
>  >  =
>  >  >  >  >  >> '0xdfb472c0' poll command abort successfully
>  >  >  >  >  >> [ 9074.106547] arcmsr0: abort device command of scsi id = 0 lun
>  >  = 0
>  >  >  >  >  >>
>  >  >  >  >  >> This is my setup:
>  >  >  >  >  >>
>  >  >  >  >  >> 1 x MSI K8N Master2-FAR
>  >  >  >  >  >> 1 x Opteron 252
>  >  >  >  >  >> 1 x Areca ARC1200 (sitting in a PCIe x4 socket)
>  >  >  >  >  >> 2 x WD1500ADFD in RAID1
>  >  >  >  >  >>
>  >  >  >  >  >> astan@rubik:~$ uname -a
>  >  >  >  >  >> Linux rubik 2.6.24-7-generic #1 SMP Thu Feb 7 01:29:58 UTC 2008
>  >  >  i686
>  >  >  >  >  >> GNU/Linux
>  >  >  >  >  >> astan@rubik:~$ modinfo arcmsr
>  >  >  >  >  >> filename:
>  >  >  >  >  >> /lib/modules/2.6.24-7-generic/kernel/drivers/scsi/arcmsr/arcmsr.
>  >  ko
>  >  >  >  >  >> version:        Driver Version 1.20.00.15 2007/08/30
>  >  >  >  >  >> license:        Dual BSD/GPL
>  >  >  >  >  >> description:    ARECA (ARC11xx/12xx/13xx/16xx) SATA/SAS RAID
>  >  HOST
>  >  >  Adapter
>  >  >  >  >  >> author:         Erich Chen <support@xxxxxxxxxxxx>
>  >  >  >  >  >> srcversion:     28EAD6AB49D4491CA04D465
>  >  >  >  >  >> [...]
>  >  >  >  >  >>
>  >  >  >  >  >> I've read some previous posts here on LKML that it could be the
>  >  >  Areca
>  >  >  >  >  >> firmware who doesn't like my WD disks. Anyone know if this is an
>  >  >  IRQ
>  >  >  >  >  >> handling problem in the kernel, or if it's a problem with the
>  >  RAID
>  >  >  >  >  >> controller firmware?
>  >  >  >  >  >>
>  >  >  >  >  >> Erich Chen (of Areca); have you tried the new ARC1200 in RAID1
>  >  >  >  >  >> configuration with Raptor disks on Linux?
>  >  >  >  >  >>
>  >  >  >  >  >> As a side note, I can tell you that I first tried running
>  >  FreeBSD
>  >  >  6.3
>  >  >  >  >  >> (RELENG_6) on this machine, but got random reboots during disk
>  >  I/O
>  >  >  >  >  >> (even with a kernel with KDB debugging turned on). This leads me
>  >  to
>  >  >  >  >  >> believe that it might be a firmware issue, and that Linux just
>  >  >  handles
>  >  >  >  >  >> it more gracefully than FreeBSD.
>  >  >  >  >  >>
>  >  >  >  >  >> Any ideas or advice is appriciated. This is my first post to the
>  >  >  LKML,
>  >  >  >  >  >> so please instruct me if you want more information or if you
>  >  want
>  >  >  me
>  >  >  >  >  >> to take further debugging actions.
>  >  >  >  >  >>
>  >  >  >  >  >> Best regards,
>  >  >  >  >  >> Aron Stansvik
>  >  >  >  >  >
>  >  >  >  >
>  >  >  >  >
>  >  >  >
>  >  >
>  >  >
>  >
>  >
>
N????y??b???v???{.n???{??zX???}????j:+v????zZ+€?zf??????i???????????f?^j?y??@A???0?h??i