Hi folks, I have just installed a PCI SATA controller based on the SIL3112A chipset into my Centos-3 (latest kernel patches applied) system and moved an existing (not the boot disk) SATA drive from the onboard controller to the SIL-based one so I can setup software RAID with two 300GB drives. Kudzu/Linux found the controller and drives straight away and they've appeared as 'hda' and 'hdc'. I've managed to partition hdc (hda was already setup) and format it - this was VERY slow and 'top' showed a significant amount of interrupt time. A bit of Googling found the right command to enable DMA (hdparm -X66 -d1 /dev/hda etc.) and this seems to have cured that problem! While Googling I found a LOT of info on the SIL controller and Linux but it was all for various kernels and linux versions - can anyone steer me in the right direction for info on this setup with RHEL3/Centos - specifically: 1) Should I just use the native SATA drivers like I'm doing now or are there better ones to use? 2) Are there any drivers that will use the BIOS-based software RAID or is using the Linux software RAID the way to go? 3) Tweaks and recommendations for the setup would be appreciated. The system is going to be used as a backup server for bacula with various clients dumping data via 512K broadband and so the performance of the RAID system does not need to be blazingly fast - hence the I'm seeing if the cheap (?14!) controller is Ok for the job before trying anything else. Any info or pointers appreciated. Thanks Nigel Kendrick
On Thu, Feb 03, 2005 at 11:08:37AM -0000, Nigel Kendrick wrote:> Hi folks, > > I have just installed a PCI SATA controller based on the SIL3112A > chipset into my Centos-3 (latest kernel patches applied) system and > moved an existing (not the boot disk) SATA drive from the onboard > controller to the SIL-based one so I can setup software RAID with two > 300GB drives. Kudzu/Linux found the controller and drives straight away > and they''ve appeared as ''hda'' and ''hdc''. > > I''ve managed to partition hdc (hda was already setup) and format it - > this was VERY slow and ''top'' showed a significant amount of interrupt > time. A bit of Googling found the right command to enable DMA (hdparm > -X66 -d1 /dev/hda etc.) and this seems to have cured that problem!You have to use sata_sil module (libata) but if you are lucky like me, you will not have a luck with performace. It heavily depends on HW vendor (not on chipset itself). You may try to use kernel 2.6 too.> The system is going to be used as a backup server for bacula with > various clients dumping data via 512K broadband and so the performance > of the RAID system does not need to be blazingly fast - hence the I''m > seeing if the cheap (?14!) controller is Ok for the job before trying > anything else.Better to switch to Intel chipset SATA interface or to IDE ATA. -- Milan Kerslager E-mail: milan.kerslager@pslib.cz WWW: http://www.pslib.cz/ke/
Milan Ker?l?ger wrote:>On Thu, Feb 03, 2005 at 11:08:37AM -0000, Nigel Kendrick wrote: > > >>Hi folks, >> >>I have just installed a PCI SATA controller based on the SIL3112A >>chipset into my Centos-3 (latest kernel patches applied) system and >>moved an existing (not the boot disk) SATA drive from the onboard >>controller to the SIL-based one so I can setup software RAID with two >>300GB drives. Kudzu/Linux found the controller and drives straight away >>and they''ve appeared as ''hda'' and ''hdc''. >> >>I''ve managed to partition hdc (hda was already setup) and format it - >>this was VERY slow and ''top'' showed a significant amount of interrupt >>time. A bit of Googling found the right command to enable DMA (hdparm >>-X66 -d1 /dev/hda etc.) and this seems to have cured that problem! >> >> > >You have to use sata_sil module (libata) but if you are lucky like me, >you will not have a luck with performace. It heavily depends on HW >vendor (not on chipset itself). > >You may try to use kernel 2.6 too. > > >A better answer is to avoid using this "raid" chipset entirely and treat it as a dumb SATA IDE controller. Then use the linux software raid tools which are probably a lot better tested and offer outstanding performance. Cheers, -- Chris Mauritz chrism@imntv.com VP & chief Technology Officer Independent Music Network, Inc. http://www.imntv.com -- No virus found in this outgoing message. Checked by AVG Anti-Virus. Version: 7.0.300 / Virus Database: 266.6.0 - Release Date: 3/2/2005