Originally Posted by StefanJ
Disk /dev/sda: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Number Start End Size File system Name Flags
1 34s 976773134s 976773101s RAID: LinuxRAID raid
Disk /dev/sdc: 976773168s
Sector size (logical/physical): 512B/4096B
Partition Table: msdos
Number Start End Size Type File system Flags
1 34s 976773132s 976773099s primary
The reason you're still having a problem is because you didn't change the start to 32 like I suggested.
But actually, this doesn't matter, you have too many problems here really to proceed in an ideal fashion. Is this a BIOS or UEFI computer? sda is currently GPT with a partition that starts on LBA 34, which is an old parted and fdisk starting value. I really don't think Fedora 16's fdisk or parted used that value but it's been a while since I've done an F16 install.
Further it's weird that sda's partition scheme is GPT, but the partition scheme within the array is MBR. To my knowledge, F16's anaconda produced mirrored bootable installs by partitioning the disks identically first, and then creating separate arrays from those partitions; not creating a partitionable array. So this is not a standard installation.
Next, you have swap on the mirrored array. This is not a good idea because the md driver doesn't keep swap mirrored identically, so if you do a scrub, mismatches will be found due to swap being mirrored, and that taints your results because you won't know if all of the mismatches are due to swap or if you have actual legitimate mismatches, which indicates a problem.
So I personally recommend you just do a backup (or two) of whatever you care about, and then reinstall either F17 or F18 while creating a new mirrored array. This will setup the partitioning correctly, instead of something I've helped you bandaid together that ultimately isn't ideal. Ergo, if either drive fails, you should still have a bootable system, and that can't happen now because the only value you can choose that's 4K aligned for the new disk makes the MBR gap too small to fit GRUB into, and that means if sda dies, you can't boot the computer from sdc.
Last, don't forget to check the SCT ERC of both your drives:
smartctl -l scterc /dev/sd[ac]
If disabled, you need a startup script that sets these to something like 70 deciseconds. That way the drive gives up on error correction before the linux SCSI layer times out. When the drive reports a read error, the md driver will get the data from the other drive, and most likely fix the bad sector causing the original read problem. This is a per disk setting.
If unsupported, you'll need to change the SCSI layer block device timeout. This is also per device, and needs to be set at startup time.
echo 130 >/sys/block/sdX/device/timeout
That will change it from the default of 30 seconds, to 2 minutes 10 seconds, which is at least 10 seconds longer than the 1.5 to 2 minute drive timeout of a typical consumer drive that doesn't support settable SCT ERC.