[ Date Index ][
Thread Index ]
[ <= Previous by date /
thread ]
[ Next by date /
thread => ]
On Friday 27 August 2004 07:55, Simon Waters wrote:
It is hardware RAID controller that presents SCSI, but has SATA disks behind it. cat /proc/scsi/megaraid/0 LSI Logic MegaRAID 6.62 254 commands 16 targs 7 chans 7 luns But surely that is the point of hardware RAID - the OS only needs a minimal clue that it is happening for administrative purposes.
I think the cheap IDE hardware RAID stuff does rely on drivers to do a lot of the work... The main problem is that (at least for hardware IDE RAID) the drivers are often closed source, poorly tested and dangerously outdated...
IDE RAID in software is rock solid in my experience, as much as software RAID ever is (i.e. write some crap to one disk only and wait for the crash ;-). Indeed I've always found Linux software RAID stuff remarkably resilient given the "quality hardware" usually thrown at it. How did the server die? How was it configured?
It was a Redhat 8.0 machine with an onboard Promise FastTrack IDE RAID controller doing RAID 1. One day the machine just died. Rebooted it to find that the filesystem was trashed beyond repair (thank god for backups). I was unable to do a proper post-mortem at the time (it was reinstalled with Windows 2000 before I got a chance), but I think it's reasonable to conclude that the RAID driver was to blame. I might give IDE RAID another chance and play with it on a spare machine, but I'm in no rush to use 2 year-old non-free drivers which will only work with 2.4 kernels. Perhaps it is better to stick to software RAID with IDE - I must give that a try sometime. Is there support for SATA RAID actually in the kernel or do you need to compile third-party modules? I'd be much happier using a free and well-tested driver which doesn't rely on the hardware manufacturer to maintain it. Cheers, David. -- David Johnson http://www.david-web.co.uk/ http://www.penguincomputing.co.uk/ -- The Mailing List for the Devon & Cornwall LUG Mail majordomo@xxxxxxxxxxxx with "unsubscribe list" in the message body to unsubscribe.