[ale] low level format

[Phil Turmel]

On 06/27/2012 06:04 PM, Chuck Payne wrote:
> 
> More information, smartd was reporting 4 bad sectors, then last week I
> was getting a weird error that there was a hole in the EXT in dmesg
> and the drive went missing until I reboot and next I am see this in...
> 
> [41442.498058] ata8: softreset failed (device not ready)
> [41442.498068] ata8: hard resetting link
> [41446.996085] ata8: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
> [41447.010713] ata8.00: configured for UDMA/33
> [41447.010801] ata8: EH complete
> 
> So I am hopeing that the lowlevel will fix this, if not I have a 1.5T
> to replace 1T.
> 

It might help you to know that most desktop-type hard drives on the
market have very aggressive error recovery firmware, to the extent that
they'll spend minutes attempting sector reconstruction before they'll
give up and report the read error.  This can be a problem in Linux,
especially with RAID setups, as the default Linux timeout is 30 seconds.
Then the driver tries to reset the link, the drive ignores the
controller (still desperately trying to recover the data), and you get
the above errors in your logs.

If the device is in a RAID array, that unresponsiveness will get the
drive unnecessarily bumped out the array.

You might want to increase that timeout, like so (assuming /dev/sdi):

# echo 120 >/sys/block/sdi/device/timeout

Meanwhile, it might help us offer advice on the seriousness of your
drive problem if you share the output of:

smartctl -x /dev/sdi

Many sector errors are transient--they're fixed if the offending
sector(s) are rewritten with new data.  Sectors that might be fixable
are labeled "Pending".  The write attempt will also cause
relocation if they can't be fixed.

HTH,

Phil