[GLLUG] Replacement disk in RAID

[James Dutton]

On Sun, 30 Apr 2023 at 08:07, Dr. Axel Stammler via GLLUG <
gllug at mailman.lug.org.uk> wrote:

> Good morning,
>
> one of four hds in my LVM2 raid has stopped reacting to SMART commands or
> anything else. The 16 TB raid system contains four 8 TB hds in a RAID 1 set
> of two RAID 4 pairs:
>
> RAID 1 ┌ RAID 4 ┌ A
>         │        └ B ← needs to be replaced
>         │ RAID 4 ┌ C
>         └        └ D
>
> So I need to replace hard disk B, buy a new disk and integrate it.
>
> Should I replace disk A, too, as it is of the same age? If not, should the
> replacement B be larger so that in the future I can grow the array? Should
> I get an HDD or an SSD?
>
> Any tips on the choice of hard disk (size, parameters, maker) as well as
> on the procedure would be much appreciated.
>
>
>
In general, one replacing a drive in a RAID setup with the same make/model
as the old one. If that model is not available, replace it with a similar
drive with at least the same amount of sectors. For striped drives, it is
best to keep all the disks at the same speed, so mixing SSD and HDD is not
a good combination for RAID setups.

Another thing to consider is something called "Object stores".  This is
where the redundancy is handled at the filesystem or above instead of the
physical disk.
It lets you do things like "Store each file on at least 2 different
physical disks."
An advantage to "Object stores" is you can easily mix disks of different
sizes, add new disks, replace faulty ones with ones of different size and
the "Object store" takes care of the distribution of files.
btrfs is an example of such an "Object store" filesystem, but there are
many others.

Kind Regards

James
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.lug.org.uk/pipermail/gllug/attachments/20230430/6f84b22d/attachment.htm>