Re: SATA disks resets in a md setup

Re: SATA disks resets in a md setup

[Jeff Garzik]

Vassilis Virvilis wrote:
> [ 9351.377961] ata2: SError: { PHYRdyChg PHYInt 10B8B Dispar }
> [ 9351.377983] ata2.00: cmd b0/d5:01:09:4f:c2/00:00:00:00:00/00 tag 0 pio 512 in
> [ 9351.377985]          res 50/00:00:b6:46:6a/00:00:13:00:00/e0 Emask 0x10 (ATA bus error)
[...]
> [10665.354196] ata2: SError: { UnrecovData Handshk }
> [10665.354196] ata2.00: cmd 35/00:00:27:ae:7a/00:04:01:00:00/e0 tag 0 dma 524288 out
> [10665.354196]          res 50/00:00:26:ae:7a/00:00:01:00:00/e0 Emask 0x10 (ATA bus error)
[...]
> and my filesystem is dead. /dev/sdb is deleted from /dev. I have to reboot and even then linux can't find the ata2 /dev/sdb.
> I have to remove power for 1-2 min for the disk to become accessible again.
> 
> Do you think the disk is bad or something?


For hardware details, see 
http://ata.wiki.kernel.org/index.php/Libata_error_messages

The ATA bus is the cable connection, so an ATA bus error typically means

- problem with your cable, or
- your motherboard's SATA port, or
- your drive's SATA port, or
- "dirty power" supply, or
- some other cause for cable interference

Regards,

	Jeff

Re: SATA disks resets in a md setup

[v.virvilis]

On Sat, 09 May 2009 03:35:05 -0400, Jeff Garzik <jeff@garzik.org> wrote:

> 
> For hardware details, see 
> http://ata.wiki.kernel.org/index.php/Libata_error_messages

Thanks for the link.
> 
> The ATA bus is the cable connection, so an ATA bus error typically means
> 
> - problem with your cable, or
> - your motherboard's SATA port, or
> - your drive's SATA port, or
> - "dirty power" supply, or
> - some other cause for cable interference
> 

I have changed SATA cables twice and I added a better PSU. The problem
persists.

Do you have any insight on the sectors count mismatch I mentioned in the
first mail?
The disk is 250GB but it looks it is searching for a 500GB disk (that is
md0 = sda + sdb).

Is it possible the SATA reset thing to trigger an md bug. I am totally
guessing here...

Regards
            .bill

Re: SATA disks resets in a md setup

[Vassilis Virvilis]

On Saturday 09 May 2009, Jeff Garzik wrote:
> 
> For hardware details, see 
> http://ata.wiki.kernel.org/index.php/Libata_error_messages

thanks for the link

> 
> The ATA bus is the cable connection, so an ATA bus error typically means
> 
> - problem with your cable, or
> - your motherboard's SATA port, or
> - your drive's SATA port, or
> - "dirty power" supply, or
> - some other cause for cable interference
> 

Ok I changed
	M/B,
	PSU
	and cables.

Now the stress test passes only one SATA reset instead of 3 or 4 before the fatal one.


[ 1804.915319] ata1.01: exception Emask 0x10 SAct 0x0 SErr 0x10000 action 0xe frozen
[ 1804.915319] ata1.01: ST-ATA: DRQ=1 with device error, dev_stat 0x0
[ 1804.915319] ata1: SError: { PHYRdyChg }
[ 1804.915319] ata1.01: cmd b0/d5:01:09:4f:c2/00:00:00:00:00/10 tag 0 pio 512 in
[ 1804.915319]          res 00/00:01:09:4f:c2/00:00:00:00:00/10 Emask 0x212 (ATA bus error)
[ 1804.915319] ata1: hard resetting link
[ 1810.279540] ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
[ 1810.305230] ata1.00: configured for UDMA/133
[ 1810.314698] ata1.01: configured for UDMA/133
[ 1810.314698] ata1: EH complete
[ 1810.318713] sd 0:0:0:0: [sda] 488397168 512-byte hardware sectors (250059 MB)
[ 1810.318713] sd 0:0:0:0: [sda] Write Protect is off
[ 1810.318713] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
[ 1810.318713] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1810.322654] sd 0:0:1:0: [sdb] 488397168 512-byte hardware sectors (250059 MB)
[ 1810.326655] sd 0:0:1:0: [sdb] Write Protect is off
[ 1810.326655] sd 0:0:1:0: [sdb] Mode Sense: 00 3a 00 00
[ 1810.326655] sd 0:0:1:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1810.330758] sd 0:0:0:0: [sda] 488397168 512-byte hardware sectors (250059 MB)
[ 1810.330758] sd 0:0:0:0: [sda] Write Protect is off
[ 1810.330758] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
[ 1810.330758] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1810.334656] sd 0:0:1:0: [sdb] 488397168 512-byte hardware sectors (250059 MB)
[ 1810.334656] sd 0:0:1:0: [sdb] Write Protect is off
[ 1810.334656] sd 0:0:1:0: [sdb] Mode Sense: 00 3a 00 00
[ 1810.334656] sd 0:0:1:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA

 Regards

   .bill

Re: SATA disks resets in a md setup

[Tejun Heo]

Vassilis Virvilis wrote:
> Ok I changed
> 	M/B,
> 	PSU
> 	and cables.
> 
> Now the stress test passes only one SATA reset instead of 3 or 4 before the fatal one.
> 
> 
> [ 1804.915319] ata1.01: exception Emask 0x10 SAct 0x0 SErr 0x10000 action 0xe frozen
> [ 1804.915319] ata1.01: ST-ATA: DRQ=1 with device error, dev_stat 0x0
> [ 1804.915319] ata1: SError: { PHYRdyChg }
> [ 1804.915319] ata1.01: cmd b0/d5:01:09:4f:c2/00:00:00:00:00/10 tag 0 pio 512 in
> [ 1804.915319]          res 00/00:01:09:4f:c2/00:00:00:00:00/10 Emask 0x212 (ATA bus error)
> [ 1804.915319] ata1: hard resetting link
> [ 1810.279540] ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 300)

PHYRdyChg under load is very symptomatic of inadequate power supply.
If you run "smartctl -a" on the device before and after the error,
what counters change?

If you have two PSUs around, one thing worth trying is to power up the
second PSU separately and put half of the drives on the separate PSU
and see whether the problem goes away or the pattern of failures
changes.  PSU can be easily powered up w/o motherboard.

  http://modtown.co.uk/mt/article2.php?id=psumod

-- 
tejun