The problem with the RAID array being hosed when you create a file system has a simple explination. HPT raid controllers are normally termed WinRAID devices; they rely on specialised device drivers to provide the RAID functionality (as against true RAID controllers which use embedded software). On Linux the device drivers are provided by the ataraid and hptraid modules.
The HPT RAID controllers store information on the RAID disks that describe the configuration of the RAID array. Because Highpoint designed the controllers for PCs, they assumed that Windows would be the only OS that they would have to worry about, and therefore put the RAID configuration information on disk sectors that are normally unused by MS file systems (FAT, NTFS, etc). However these sectors are typically used by Linux filesystems (ext2/3, JFS, ReiserFS, XFS), so when the file system is created on the RAID array it trashes the configuration information. Hence the complaints from the RAID BIOS. The ataraid and hptraid modules are really intended to allow you to access MS file systems on an existing RAID array, not to create your own file systems. I have a HPT-370 controller (basically the same as the 372, but only supports UDMA-100 not -133) and I've found that the best way to use it is to disable the RAID configuration (i.e. use it as a normal IDE controller) and use the Linux software raid (md) system. Its at least as fast as the ataraid/hptraid combination. Hope the above helps, even if it is long-winded. ----- Original Message ----- From: <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Friday, September 13, 2002 7:59 AM Subject: Re: [Cooker] HPT372 raid controller not supported? On Thursday 12 September 2002 03:33 am, Stéphane Teletchéa wrote: > Due to no answer i my last message (sorry i couldn't go through the betas > stages, the hardware is 4 days old), if flipped connectors (nappes IDE - > please translate) 40/80, the harddrive, and finally found a pseudo-working > configuration : > I See the same errors if I dont use ide=nodma > > Before the hd in hdb was in hdg (reiserfs partitions) and produced the same > errors messages, so i presume it is a not-yet supported chipset ? Actually I didnt catch if you had raid enabled or not, but the chipset seems to be supported, my bios version 2.31 reports 370/372 and the "open source" driver from www.highpoint-tech.com is the same for the 370 and 372. I can make partitions on the raid array during an install (ataraid.o and hptraid.o loaded) but for some reason the diskdrake writes the partition table to the drive in a way that hoses the Raid array. upon reboot HPT bios complains the raid array is broken and you have to re-create it and lose everything on the drive. After installing on a regular drive I can load ataraid and hptraid and access the partitions on the raid array, but it is choppy and slow. I can not use diskdrake. During an install the raid partitions show up as /dev/ataraid/d0p? which diskdrake likes, however when loading the modules in a running system the devices come up as /dev/ataraid/disc0/part? which diskdrake does not recognize, making a link from /dev/ataraid/disc0/part? to /dev/ataraid/d0p? lets diskdrake work. I had the /, /home, and swap on the raid0 array in 8.2 and it worked beautifully, so far the results in 9.0 are dissappointing, but at least built in support is there. Building support into the kernel and disabling the other ide stuff you dont need may take care of the sluugish response. But thrashing the raid array is a BAD thing :(
