[Cialug] lost+found

[Scott Yates]

Yes, but what i am suggesting is setting up a spare server with the new
drives and making it the backup server.  Once you get a good FULL backup of
the clients, THEN diagnose/fix the old server.  If as you said, you need to
keep maximum up-time, then a spare server is basically a minimum
requirement.   Drives fail.  Sometimes obviously, sometimes not.  As you
get over a certain amount of storage, it is statistically almost guaranteed
(something like 99.8% certain) that you will have bit rot and unrecoverable
reads.  You just try to plan for it as best as your budget allows, and make
sure the people in charge are aware of the risks.

You might also check out Amazon's new "slow storage" i think they call it.
 Basically it gets you cloud storage VERY cheap if you are flexible about
how quickly it is retrievable.



On Sun, Dec 16, 2012 at 11:16 PM, L. V. Lammert <lvl at omnitec.net> wrote:

> On Sun, 16 Dec 2012, Scott Yates wrote:
>
> > 13x 3tb drives would get you a raid-5 that should cover the storage while
> > you diagnose.  You could expand the storage after diagnosing the problem,
> > or keep them online for a hot spare server.  The investment in drives is
> > probably more than offset by the data you are backing up.
> >
> Of course, but, in practicality, quite impossible. It would take over five
> days on GBE each way, total 10 days downtime.xcx
>
> I realized long ago that when deailing with these huge arrays, you have to
> rely on redundant HW (RAID6 in this case), as moving data around is
> just not practical.
>
> `       Thanks!
>
>         Lee
> _______________________________________________
> Cialug mailing list
> Cialug at cialug.org
> http://cialug.org/mailman/listinfo/cialug
>