| 1 |
Benoit St-Pierre schrieb: |
| 2 |
> I'm in the planning stages of setting up a file server and am considering |
| 3 |
> using RAID. |
| 4 |
> |
| 5 |
> My concern is that my drive sizes are mixed. I have two 500GB SATA drives, a |
| 6 |
> 320GB IDE and a 250GB IDE. |
| 7 |
> |
| 8 |
> I would like to set these up so that the maximum amount of disk space is |
| 9 |
> usable, but still be able to recover from any one drive failing. I would |
| 10 |
> also like to be able to add drives of any size as easily as possible. |
| 11 |
> |
| 12 |
|
| 13 |
I can think of two ways: |
| 14 |
|
| 15 |
1. RAID1 over those 2 SATAs + RAID1 over the IDEs, then an LVM on top of |
| 16 |
both. |
| 17 |
This wastes 70GB and uses a total of 750GB for redundancy. New disks can |
| 18 |
be added in increments of two (forming a new RAID1 which is then added |
| 19 |
to the LVM volume group) |
| 20 |
|
| 21 |
2. Linear or RAID0-arrangement over the two IDEs, RAID5 over this RAID |
| 22 |
and the other two disks. |
| 23 |
This wastes 70GB, too and uses a total of 500GB for redundancy. |
| 24 |
|
| 25 |
Note1: NEVER EVER build some kind of RAID other than "Linear" (also |
| 26 |
called JBOD) over two IDE disks on the same cable. Performance will |
| 27 |
suffer greatly as will security because most simple onboard controllers |
| 28 |
can't handle a dying disk and that one might take the other one with it |
| 29 |
into death. |
| 30 |
|
| 31 |
Note2: RAID-autodetection doesn't always work with RAIDs over RAIDs. It |
| 32 |
is better to deactivate RAID-autodetection and tell the kernel directly |
| 33 |
which devices shall be created in which order. See: |
| 34 |
/usr/src/linux/Documentation/md.txt |
Archives