Raidframe low I/O for parity rebuild, but when accessed through rraid* devices

[Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost>]

Hi list,

I am posting the odd behaviour I am currently experiencing with 
Raidframe on a freshly installed system (amd64, NetBSD-6.1), in the hope 
that someone encountered this already.

Just after initializing the raid 1 array using two ATA disks (1 SATA, 1 
PATA), starting the parity rebuild leads to a curious behaviour: when 
monitoring the disks I/Os via iostat, the first few minutes of the 
parity rebuild go fine, both disks working at about 20-25MiB/s (which is 
their nominal speed when used in !raid setup).

But after that, the rebuild I/Os slow to a crawl to a mere 2-3MiB/s and 
it remains stationary. However, accessing the raid (formatting, copying 
files, dd'ing...) does not seem to be affected by this slowness, as the 
disks do work at their nominal speed when properly solicited. It is just 
as if parity rebuild decided to restrict its bandwidth or priority 
(although no other bg activity happens on those disks ATM).

Does this ring a bell to someone?

That is a minor annoyance as the raid set is working, but the parity ETA 
is expected to take a full day, instead of an hour or so...

FWIW the hardware is kinda old: 250GiB SATA-I disk + 250GiB PATA/IDE 
disk in raid1 configuration (512b sectors, partitions are aligned on 
1MiB boundary). Disks have low I/O capabilities (about 25MiB/s each), 
controllers are in AHCI. raidframe config is following (no big surprise 
here):

START array
1 2 0

START disks
/dev/wd0e
/dev/wd2e

START layout
128 1 1 1

START queue
fifo 100

Tried swapping different disks, modifying the layout, or using SATA-only 
and PATA-only disks, but it always ends up that way. Next on the list is 
to test with -current, but I need to install radeon DRM firmwares first (:

Hints are, of course, appreciated.

Cheers,

--
Jean-Yves Migeon