0

Are partition-tables from hard-drives, that are part of a RAID (no matter of what kind of RAID) always equal to the partition-table configured for the RAID?

What is clear: A machine has a RAID 1 with four hard-drives. When I shutdown the machine and remove one hard-drive after the other to read out the partition-table of each hard-drive, every hard-drive has the same partition-table as configured for the RAID 1. This makes sense ... The same result I had with RAID 1+0 and RAID 1e.

What is unclear: What kind of results I have to expect from hard-drives that are part of other RAIDs like RAID 0 or RAID 5?

edi38
  • 1
  • 1

2 Answers2

2

Depending on the type of RAID, it's entirely possible to end up with a disk without any partition table at all. Hardware RAID will have its own information even below OS structures like the partition table and you can end up with many disks without a copy of this table.

With software RAID, you often have RAID partitions that just declare "this thing is managed by the RAID system, but it's entirely possible to add a "raw disk" without a partition table to a RAID.

In any case, single disks out of a RAID 0 or RAID 5+ array will never have complete information. Kill 1 disk in a RAID 0, 2 in RAID5 or 3 in RAID6 and you have guaranteed data loss.

Sven
  • 98,649
  • 14
  • 180
  • 226
  • There is no concept or standard, that defines where to put the partition-table. So the controller (no matter hard- or software RAID) - or better the vendor of the controller - decide, whether a partition-table is on a RAID disk or not. Is that right? – edi38 Aug 15 '16 at 13:47
  • There *are* standards where to put the partition table. It's just that a hardware RAID controller hides fact that the "disk" is really 24 disks in a RAID6 array away from the OS. The OS sees one big disk and creates a partition table on it but what the RAID controller actually writes is something more complicated. In case of a RAID6 with x disks, the partition table gets physically written onto one disk and 2 variants get written in some redundancy format that can be used to reconstruct the data in case of disk failure. – Sven Aug 15 '16 at 15:11
-1

Raid 1 is mirroring so you get the same on both sides. Raid 0 is odd numbered blocks of data on one disk (1 3 5 7 ....) and enen blocks on the other (2 4 6 8 ...)

So no you will not get the same data on all disks.

Allmost no one uses raid 2 3 and 4.

raid 5 gives you odd and even data blocks on at least 2 disk and checksum on an addidtional disk distributed over all disks.

raid 6 is the same with 2 differently computed checksums so you need at least 4 disks.

with raid 5 you have n+1 disks, one can die without loosing your data.

with raid six you need n+2 disks and two disks can die before you loose your data.

Uwe Burger
  • 166
  • 3