[mythtv-users] Mirroring root partition on SSD to a hard drive -- bad idea?

[John Morris]

On Thu, 2016-01-14 at 09:48 -0600, Craig Huff wrote:
> I have migrated my system disk to an SSD with /var mounted on a separate
> hard drive to limit wear on the SSD. For protection against the day the SSD
> fails, I'm thinking of mirroring it to a partition on an available hard
> drive.  It appears I can do this with either the traditional software raid
> approach using mdadm that looks a lot like what I did with Sun Solaris
> systems many years ago, or with dmraid, taking advantage of the Intel
> hardware (firmware?) support on my motherboard, although that may be an
> issue with the SSD on a SATA 3.0 port and the hard drive on a SATA 2.0
> port.  If I go ahead, I'll have to investigate that.

My advice?  Don't.  I used to worry about SSD service life but I have
now tried it in production.  Have an Evergreen library ILS system here
that is running entirely on SSDs, Postgresql server included, and has
been going 24/7 since 2010.  SMART says the wear level indicator has
dropped from 100 to 97 during those years.  Granted these are good (ok,
good in 2010) Intel X25-M drives bought in rack mount servers so bargain
bin units might not be quite as reliable but they have all probably
gotten better too, nature of tech.

Of course I'm paranoid, they are all RAID1 just in case; because failure
isn't an option on these units.  I assumed at the time that there would
be failures from the constant writes but heck, get a few years out of
them and replace with something newer and bigger and enjoy the speed
while they last.  Nope, no failures yet.  So don't stop making backups
(and remember: RAID is not backup!) but go ahead and run your system on
SSD, you will love the performance.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: This is a digitally signed message part
URL: <http://lists.mythtv.org/pipermail/mythtv-users/attachments/20160114/360eee88/attachment.sig>