Re: Data corruption on software raid.

[Bill Davidsen <davidsen@tmr.com>]

Justin Piszcz wrote:
>
>
> On Sun, 18 Mar 2007, Sander Smeenk wrote:
>
>> Hello!  Long story. Get some coke.
>>
>> I'm having an odd problem with using software raid on two Western
>> Digital disks type WD2500JD-00F (250gb) connected to a Silicon Image
>> Sil3112 PCI SATA conroller running with Linux 2.6.20, mdadm 2.5.6
>
> [[ .. snip .. ]]
>
> See comments below.
>
> | Personalities : [raid1]
> | md0 : active raid1 hda2[0] hdb1[1]
> |       120060736 blocks [2/2] [UU]
Your comments below are thoughts on the PATA RAID1, not the one which is 
giving trouble, the SATA. Unless you think that making the working array 
faster, perhaps he's looking for ideas on fixing the array which isn't 
working as expected.
>
> My main question:
>
> Why the 'heck' are you running a RAID1 with a master/slave 
> combination? That is probably the -worst- way to run it.  When using 
> any form of RAID, make sure you do not share an IDE channel for any 
> one raid device.
>
> Advice?  Hook up each drive as a master, then your rebuild speed 
> should go to 30-60MB/s.
>
> Traditional Troubleshooting:
>
> What does fdisk -l /dev/hda
>           fdisk -l /dev/hdb
>
> Report?
>
> Also,
>
> What does:
>
> smartctl -a /dev/hda
> smartctl -a /dev/hdb
>
> show?
>
> Then,
>
> smartctl -t short /dev/hda
> smartctl -t short /dev/hdb
>
> Wait 5-10 minutes, re-run the commands (-a) above.
>
> Then,
>
> smartctl -t long /dev/hda
> smartctl -t long /dev/hdb
>
> Then re-run the (-a) smartctl above.
>
> --------
>
> Those errors look really weird, I would separate the two disks, each 
> on their own IDE channel and see if your problem goes away.
>
> Justin.

-- 
bill davidsen <davidsen@tmr.com>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979