Screwiness with PCI-E SATA card?

[Jonathan Gardner]

On 12/16/2008 06:34 AM, Todd Denniston wrote:
> Jonathan Gardner wrote, On 12/15/2008 09:52 PM:
>    
>> I've got a system with two SATA connectors that I've hooked up two
>> SATA drives to. I upgraded by adding a PCI-E SATA card and another,
>> 500GB SATA drive.
>>
>> I've got Fedora 10 running with the root partition on an LVM volume
>> group on the big 500GB drive and my /home partition on an LVM VG on a
>> RAID1 array of the two original drives. Things go fine for about an
>> hour until the root partition begins to corrupt.
>>
>> When I reboot, fsck reports that the root partition has a lot of
>> errors. Running fsck manually confirms that there are a lot of errors.
>>
>> smartctl doesn't report anything obvious, but I don't have a lot of
>> experience with smartctl.
>>
>> Does anyone have any suggestions on how to investigate this? I'm
>> rather amateurish with the admin tools to diagnose problems with
>> hardware.
>>
>> Right now I've got my distribute setup on the two original drives
>> because I can't trust the new, large drive running on the SATA card.
>>
>>      
>
> which file system?
> if ext[23] then read up on e2fsck with the following options:
> e2fsck -k -c -c -f -y -C 0 -v /dev/device
> Assumptions: 1) if ext3, the journal is internal. 2) bad blocks be the problem
> even on a new drive (been there got the t-shirt_S_). 3) /dev/device is not
> mounted at the time of execution.
>
> This should force e2fsck to check for bad blocks while doing the file system
> fixes.
>
>    

Thanks for the pointers. I am running the e2fsck command now.

Someone may find the following useful. Note that the ext3 partition is 
living on VolGroup01/LogVol02. VolGroup lives entirely on /dev/sdc2, 
which is a massive partition on the 500GB drive connected over the SATA 
PCI-E card. So the device I am using is:

     # e2fsck -k -c -c -f -y -C 0 -v /dev/VolGroup01/LogVol02
     e2fsck 1.41.3 (12-Oct-2008)
     Checking for bad blocks (non-destructive read-write test)
     Testing with random pattern:   9.07% done, 4:08 elapsed

The partition is about 20GB, and even though the drive can get some very 
fast speeds, it is going to take about an hour to run through it. I'll 
post the results here in case someone may find use of it.