[mythtv-users] Re: A warning about Samsung HDDs

[Tony Lill]

Dewey Smolka <dsmolka at gmail.com> writes:

> As someone who has professionally destroyed hard drives (most of us
> here are only practiced amateurs), what kind of recommendations/ best
> practices can you offer to ensure maximum HDD life? What are some
> warning signs we should look for, other than 'my drive has come to a
> literal screeching halt'? SMART is nifty, but in my experience has
> only warned me of problems after I knew there were problems. What can
> I do to detect a drive failure before it occurs, and to get my data
> off before the drive becomes unreadable?

First, a question. What's the current philosophy on spinning down
drives? Is it better or worse for longevity that keeping them spinning
and idle? I've got my recordings spread across a bunch of drives, 5
are exclusively used by myth, with a spin-down time of about 20
minutes. Given the typical myth usage, these can stay spun down for hours.

Now for some answers.

I've had 2 out of 3 Samsung harddrives die on me last year, both
within about 6 months of being installed. These were slow deaths, and
I got a lot of read and write errors logged, and so was able to
recover most of my data before they completely packed it in. This
permanently put me off using LVM without RAID or mirroring to back it
up.

First thing is that you should monitor the logs in /var/log. Use
logcheck or logwatch, which should both be standard packages on most
recent OS's, and actually READ the mail they send. Any harddrive
messages like read or write errors, dma errors, ide resets should send
up a red flag.

If you are smart and not using LVM, then if the filesystem is still
healthy enough to mount, you can just use tar or cpio or cp -R to get
your stuff off the drive and on to a replacement.

If not, then dd_rescue is your best friend. It's designed to try and
extract as much data off of a broken drive as possible, and copy it
block by block to a new drive. Knoppix has dd_rescue on it, so unless
you have a spare machine to do the recover on, you'll need a copy of
that on a CD, just in case. Actually, everyone should have a copy of
Knoppix, just in case.

You need to get a new drive at least as big as the old one, and
partition is so each slice you want to recover is at least as big as
the old slices. Be careful if you buy the "same" size drive from a
different manufacturer. You can wind up a a few Mb short, in which
case you're boned if you're using LVM. If you're not getting exactly
the same replacement drive, go for a bigger one.

Put the broken and replacement drive in the same machine, and use
dd_rescue to copy the data, and when it's done, fsck the results. If
you're lucky, you'll have a working filesystem without too many
missing or corrupt files. If not, well, you've been making regular
backups of everything important, right?

If your replacement drive is much larger than the original, you can
make the partitions much larger, and use your filessytem tools to
extend the filesystems into the new space.

If you are using LVM, your only option is dd_rescue, since what's on
the disk is just part of the filesystem data, and has no independent
existence. You need to replace it as exactly as possible. 

If you created the PV on the whole disk instead of a partition, I
don't know if you can add any extra space on the replacement disk to
the PV, so if you get a larger disk, the extra space may be wasted. If
you used a partition, you can always add a new partition to your LVM
after you're done.

Once you finished with using dd_rescue to recover what data you can,
you need to replace the broken drive with the new drive, bring the
logical volume up, and then run fsck, and pray.
--
Tony Lill,                         Tony.Lill at AJLC.Waterloo.ON.CA
President, A. J. Lill Consultants        fax/data (519) 650 3571
539 Grand Valley Dr., Cambridge, Ont. N3H 2S2     (519) 241 2461
--------------- http://www.ajlc.waterloo.on.ca/ ----------------
"Welcome to All Things UNIX, where if it's not UNIX, it's CRAP!"