Re: Adding RAID 6 support to the Installation Disk
Dr. Harold K. Brown, P.E. wrote:
Currently, I am using the ARC-1220 Areca raid controller with the RAID
6 engine under Debian. The system is i386 based and the only hard
drives on the system is the Raid 6 drive set. This is a hardware raid
solution.
I believe it would be beneficial for Etch users to have a Raid 6
installation option that really works at time of installation. I can
help with testing and even do some patches. My time is limited, and I
do know time is short with the upcoming planned Etch stable release.
If the installation team is open for this, please contact me and let
me know the best way to approach this if there is a desire to include
this capability into the installation disk.
This would require an addition of a device driver with a risk factor
that I believe to be low. Currently the drive is open source and has
been included into the Git Tree at kernel.org. The driver is
supported by the company now and complies with the kernel code writing
standards. I believe the drivers will be included in the 2.6.19 or
2.6.20 kernel.org. However, I understand that Etch will use 2.6.16 or
2.6.18 and thus the driver will not be included.
This may be one of the best Raid 6 solutions on the market at this
time and it does seem to work well with Debian. Another Areca
customer made a Debian install disk based on Sarge, which I have been
using. It works well, but does not have the driver support for other
common devices like sound that are in the newer kernels.
The install disk used the 2.6.8 kernel. I have used the
manufacturer's provided drivers on kernels 2.6.8, 2.6.16 (Debian
version), and 2.6.18 (kernel.org version). I used the drivers from
the git Tree and personally installed the drivers in 2.6.16 deb source
and 2.6.18 kernel.org source so that I could configure the drivers
though make menuconfig. The drivers are installed as modules. Both
kernels seem to work fine.
Note, that the purpose of this system configuration is to be 100% RAID
6 where in such a case it would be very convenient to have RAID 6
support at time of installation. Test data so far seems to support
that this may very well be possible, at least for the test hardware
set under use.
Some Functional Test Results Follow:
The following outlines one of the experiments that I performed. Keep
in mind that the only hard drives on the system is the RAID 6 set. I
configured 4 drives using the Raid's boot bios software. Once the
RAID 6 was configured, Debian was installed from CD using a modified
Sarge installation disk. During the actual install, while loading the
base system I randomly pulled one of the 4 drives from the system,
then in a few minutes pulled another. The alarms sounded, but the
install proceeded. After a few minutes I put the drives back into the
system. Keep in mind that I am proceeding with the installation
without any delay while the two drives are pulled out and then put
back into the system. Once the drives re-synced, I did it again and
again with differing drives.
I then: 1) Added the driver to the stock 2.6.18 kernel from
kernel.org, 2) With the 2.6.18 kernel loaded, did a distribution
upgrade from Sarge to Etch (Testing). 3) Then modified the Debian
2.6.16 kernel source and installed. All of this work was done on the
test system with the RAID 6 set as the only hard drive. I did
numerous demonstrations allowing any that came by to randomly select
one or two drives and pull them out at will.
The system is still in use and under test. Results are good.
I asked on the kernel list, some time ago, before the last stable
release, if the Areca driver could be included before the stable release
so as to be installable, etc.. The response at the time was if it was
included in an upstream kernel it might be able to be backported and
included in the stable release kernel. At the time the Areca driver had
not been accepted (I thought it had but it had not). Now it has been
(included in 2.6.19 change log). So it might be possible to get it
included in the 2.6.18 for the next stable.
I, for one, would very much welcome this. I am not a kernel beater but
I would be willing to do what I can to help test/whatever to see this
happen.
Thanks
R.Parr, RHCE, Temporal Arts
Reply to: