Date: Tue, 19 Aug 2003 10:51:13 +0100 From: "White, NGH (Nick) " <N.G.H.White@rl.ac.uk> To: "'aic7xxx@freebsd.org'" <aic7xxx@freebsd.org> Subject: SCSI - IDE RAID experience???? Message-ID: <350DC7048372D31197F200902773DF4C0396459B@exchange11.rl.ac.uk>
next in thread | raw e-mail | index | archive | help
I inherited a number of systems that use Fibrenetix Zero-D X-3i IDE to SCSI RAID arrays. These are based on Accusys RAID components as far as I can tell. Laterly, we have purchased some Infortrend IDE to SCSI RAID arrays. These are Infortrend 6300 arrays, which use the 7250 controller. Both these array types have 12 disks, configured as a RAID 5. The Zero-D present 800GB and the Infortrend 1.9TB after formatting. The disk drives are IDE and the firmware in both cases presents a single SCSI target to systems based on an Adaptec 39160 HBA. Over the past 2 months we have had a series of issues which are reported by the aic7xxx driver. The short list is: 1. sense key Medium Error - with additional sense indicating unrecovered read error 2.The aic7xxx driver going through the recovery routine with 'Attempting to queue an ABORT message' and 'Attempting to queue a TARGET RESET message' These invariable come back with: 'aic7xxx_abort returns 0x2002' and 'aic7xxx_dev_reset returns 0x2002' also plenty of "Dumping card state" 3. The aic7xxx driver reporting: 'parity error detected in Data-in phase', 'CRC Value Mismatch' 'Unexpected busfree in Message-out phase' and other variations. The Infortrend arrays are connected to a dual P4 2.4GHz SuperMicro motherboard with hyperthreading on and the system has 1GB of RAM. The Adaptec 39160 is implemented via a riser card in the machine. These arrays use Maxtor 200GB Diamond Max Plus 9 drives. The Infortrend based systems are using Red Hat 7.3 with the aic7xxx driver at version 6.2.8 and the kernel is a Red Hat 2.4.20-18.7smp variety. The Fibrenetix arrays are connected to dual P3 1.2GHz Supermicro motherboard machines and have 1GB of RAM. The machines contain a single Adaptec 39160. These arrays use Maxtor 80GB Diamond Max Plus 9 drives. The Fibrenetix based systems are using Red Hat 7.2 with the aic7xxx driver at version 6.2.4 and the kernel is a Red Hat customised(due to other issues) 2.4.9-31. The past 2 weeks have been particular traumaic with every day having at least one system with issues. It is also getting embarassing with our apologies to our end users. My gut feeling is that if everything is working fine then the SCSI drive emulation in the array firmware copes well with the Adaptec HBA and Linux but if one of the IDE drives has a retry or several retries, then the firmware emulation is slightly weak and the aic7xxx driver and SCSI layers in the kernel believe that we have a timeout and the driver goes into recovery mode....etc..... If there was a SCSI signalling problem then the array can detect that. We just cannot believe that we have bad SCSI cables over 2 ranges of systems bought 12 months apart but exhibit similar fault conditions. The suppliers of both arrays do not have that much Linux experience and have asked us to increase the device driver timeouts and retry count, which apparently can be done in Windows. Does anyone have any experience with SCSI to IDE RAID arrays such as the Infortrend and Fibrenetix, and hence any advice on how to provide proof that these devices maybe outside SCSI spec or help me sort out the issues?????? TIA, - Nick
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?350DC7048372D31197F200902773DF4C0396459B>