Drive I/O error in software RAID10 when assembled in KVM using virtio

[Eric Shubert]

On 02/07/2013 10:10 PM, James Dugger wrote:
> I have a software RAID10 array (4-1TB drives) that I have assembled in a
> kvm vm on Proxmox (Debian Squeeze 6.0).  The HDDs were mapped to the vm
> using virtio.  The array once created and started in the vm (during
> resync) started throwing an I/O error on one of the disks.  It kept
> repeating the error until I stopped the array.  At the host level I ran
> smartctl on all drives and they all passed inspection.
>
> The drives are not all the same brand.  The array is made up of 1
> Seagate Barracuda, 2 WD Blacks, and 1 Hitachi.  The Seagate is an
> Advanced Format with sector size  (logical/physical) equal to
> 512B/4096B.  The other drives are standard format 512B/512B.  All drives
> are configured as GPT and have been partioned (using parted) to be
> aligned as follows:
>
> Number  Start   End     Size    File system  Name         Flags
>   1      1049kB  1000GB  1000GB               udata_raid2  raid
>
> Is it possible that the physical sector size of 4096B of the Seagate
> drive is causing the problem?  I don't believe that the I/O error was on
> the Seagate (I can't recall which it was on at the time).
>
> The HDD's where mapped to virtio devices using the following:
>
> qm <vmID> virtio# /dev/disk/by-id/ata-<disk name>
>
> Any thoughts?
>
> --
> James
>

I would initially suspect the AF drive, only because it's fundamentally 
a little different. That could be an errant suspicion though.

I have a Seagate Barracuda ES.2 drive (500G, date 09404) which passes 
all tests I've thrown at it (smartctl, badblocks), yet it throws errors 
when it's put into a software raid-1. I've taken the approach which 
says, "if it doesn't work that way, don't do it that way". ;) IOW, I'm 
not using that drive with raid.

Can you identify which drive is throwing the errors?

(Hope to see you at tommorrow's meeting)

-- 
-Eric 'shubes'