On 20/08/2019 14:28, Bob Eager wrote:
On Tue, 20 Aug 2019 11:35:25 +0100, T i m wrote:

Some are up top over 50,000 hours (the Red ones).

Any idea what the MTBF is on those drives Bob?

They quote (statistical estimate) 1,000,000 hours! My experience wouls
say that's not totally improbable, given the nimber of failures I've had
over time, although I think I've only ever kept one for about 60,000
hours. That is pretty well permanently powered on in my case, low number
of spinups, and kept fairly cool (either the microservers which are well
cooled, or big Corsair cases).

How do you back that lot up or are some backups of the others?

Some back up others (on different floors of the house). Some to DVD
(using heavy ECC, see below). Important stuff goes offsite in storage
(10 miles away) and also on http://tarsnap.com

Sounds like you have most eventualities covered Bob (depending the
height above sea level of all locations and their hardness against an
EMP). ;-)

We are well above the level of the building designated as the flood
relief centre. The storage unit is a lot higher! Can't do much about the
EMP; if it covers here and North Americ as well, I doubt I'll care!

DVD ECC is via dvdisaster. I did a test; wrote a DVD with ECC then used
a Swiss Army knife to make random cuts and scrapes (took enough away
that you could see through it). It was recovered perfectly.

Wow, that sounds like the old 'Tomorrows World' demo of optical media
with jam or something. ;-)

I have a BDRW in a couple of machines and I wondered if the media was
any less reliable long term than say DVDR's? 50GB is quite a bit more
than even 9GB.

I wondered that about DVD vs. CD, and they seem pretty well the same. The
only CD failures I had were very cheap media (early days) and soem
problems with discs for about 2 years 2004-2006. I have check read back
to my oldest discs with no trouble in most cases, barring those which had
systematic problems. It's clear (I keep logs of software updates) that
one version of the writer software was glitching at the same point every
time; I suspect a buffer underrun. I managed to recover it all one way
and another as the glitch was in a metadata area. For the last two years,
everything gets ECCs added.

And they ought to survive EMP too!

I looked at the cost, usability and general hassle of backing everything
up on DVD and decided that actually a second drive was no more expensive
for the amount of data I had and the really really important stuff could
go on a SUB drive and be put somewhere safe, every so often.

You cant lose a second drive. And a 180GB of really important data is at
least 30 DVDs.

The only time I have had to fully restore is when I rebuilt the server
and upped the disk size after a primary disk went. That was a selective
restore since it had an uprated OS.

I've restored bits of the computers that are backed up on it after
crashes as well, but mostly they dont have critical data on them -
they are just 'cloud clients' onto the main server



--
The theory of Communism may be summed up in one sentence: Abolish all
private property.

Karl Marx