On 01/02/2025 at 20:45, Anton Gavriliuk wrote:
Here is raid5 setup during build time
Number Major Minor RaidDevice State
0 259 3 0 active sync /dev/nvme3n1
1 259 4 1 active sync /dev/nvme4n1
3 259 5 2 spare rebuilding /dev/nvme5n1
4 259 6 - spare /dev/nvme6n1
5 259 7 - spare /dev/nvme7n1
I'm asking because during build time according to iostat or sar,
writes occurs only on /dev/nvme5n1 and never on /dev/nvme3n1 and
/dev/nvme4n1.
So if parity is distributed in mdadm/raid5, why mdadm writes only on
/dev/nvme5n1 ?
Because, as you can see, nvme5n1 is being rebuilt from the two others.
From mdadm(1) man page:
"When creating a RAID5 array, mdadm will automatically create a degraded
array with an extra spare drive. This is because building the spare
into a degraded array is in general faster than resyncing the parity on
a non-degraded, but not clean, array. This feature can be overridden
with the --force option."
I guess this way is faster on hard disks because it implies sequential
read on N-1 drives and sequential write on the other drive, sequential
read and write speeds are similar and sequential write does not cause
significant wear. But it may differ on SSDs, spreading reads and writes
among all drives may be more efficient.
