Re: ide hardware raid
Terry Ewing wrote: Hello, I'm an administrator in a co-location facility and recently we had a customer come in to replace a raid card. The only catch was the RAID card was IDE. This was on an NT box, but I was wondering: 1) what IDE RAID cards are out there now. What is known about them. 2) What is the status of Linux support for IDE RAID cards. Thanks! - Terry Ewing [EMAIL PROTECTED] Deepwell Internet Services Hi, I'm currently checking out a couple of PCI controllers by PROMISE Technology (the FastTRAK and the FastTRAK66). The nice thing about them is that, if coupled with their "hot-swap" disk box, they allow you to remove/add a disk while the system is up and running. I couldn't find any other PCI/IDE card supporting hot-swapping. If anybody knows about them, I'd be interested too. Up to now I managed to use them succesfully on NT (:-( while when used with the current linux driver, I didn't find out how to enable the hardware-RAID. - Antonio -
HPT366 prevents boot on Abit BE-6 (probably not raid related)
I am still trying to get my HPT366 channels to work on my ABIT BE6 motherboard. I really appreciate the emails I have gotten to date, but there is still no joy. Following in the footsteps of Tim Moore, I upgraded to 2.2.13 with patches raid0145-19990824-2.2.11, and ide.2.2.13.1999.patch applied. This has made no difference. I noticed there is a debug option in hpt366.c, namely HPT366_DEBUG_DRIVE_INFO So I turned this on to see if it would help me see where the problem is. So at this point I am getting a few more messages, which may help pin down where the trouble is. The last message I now get at boot time is coming from line# 150 in hpt366.c pci_bus_clock_list: found match: 0x90c9a731 where the hex number is the value of: chipset_table-chipset_settings which is about to be returned from the subroutine: pci_bus_clock_list() This subroutine is only called from 3 places, all of which are inside the same switch statement. They all head for line 192 where there are only three statements before the next printk(), which of course never comes out. So, assuming that there in nothing sitting in a buffer someplace, (and I admit that I do not know enough about the console i/o at this stage of things to know the affects of buffering), then the system is locking up on one of those 3 statements, which are: reg2 = ~0x8000; pci_write_config_dword(HWIF(drive)-pci_dev, regtime, reg2); err = ide_config_drive_speed(drive, speed); It looks safe to rule out the 1st one. Anyone have any suggestions as to what is going wrong, or how to fix it? Best Regards, Robert Laughlin
Re: ide hardware raid
check out www.zero-d.com They make an eide internal uw scsi external raid box that looks pretty cool. -sv
HPT366 prevents boot on Abit BE-6 (probably not raid related)
Upon further investigation, I added some debug printk() statements. Assuming buffering is not confusing the issue, it looks like the machine is calling this subroutine, but the subroutine never returns; ide_config_drive_speed(drive, speed); The parms are set to these values when the call is made: drive-name = hde speed = 0x44 The subroutine ide_config_drive_speed() is part of ide.c Best Regards, Robert Laughlin -- Forwarded message -- Date: Thu, 2 Dec 1999 09:04:10 -0500 (EST) From: Robert [EMAIL PROTECTED] To: Tim Moore [EMAIL PROTECTED] Cc: raid [EMAIL PROTECTED], Andre Hedrick [EMAIL PROTECTED] Subject: HPT366 prevents boot on Abit BE-6 (probably not raid related) I am still trying to get my HPT366 channels to work on my ABIT BE6 motherboard. I really appreciate the emails I have gotten to date, but there is still no joy. Following in the footsteps of Tim Moore, I upgraded to 2.2.13 with patches raid0145-19990824-2.2.11, and ide.2.2.13.1999.patch applied. This has made no difference. I noticed there is a debug option in hpt366.c, namely HPT366_DEBUG_DRIVE_INFO So I turned this on to see if it would help me see where the problem is. So at this point I am getting a few more messages, which may help pin down where the trouble is. The last message I now get at boot time is coming from line# 150 in hpt366.c pci_bus_clock_list: found match: 0x90c9a731 where the hex number is the value of: chipset_table-chipset_settings which is about to be returned from the subroutine: pci_bus_clock_list() This subroutine is only called from 3 places, all of which are inside the same switch statement. They all head for line 192 where there are only three statements before the next printk(), which of course never comes out. So, assuming that there in nothing sitting in a buffer someplace, (and I admit that I do not know enough about the console i/o at this stage of things to know the affects of buffering), then the system is locking up on one of those 3 statements, which are: reg2 = ~0x8000; pci_write_config_dword(HWIF(drive)-pci_dev, regtime, reg2); err = ide_config_drive_speed(drive, speed); It looks safe to rule out the 1st one. Anyone have any suggestions as to what is going wrong, or how to fix it? Best Regards, Robert Laughlin
Re: ide hardware raid
RAIDZONE www.raidzone.com is a full function RAID solution thatruns under Linux (and Windows NT) using IDE (Ultra ATA) disk drives. RAIDZONE features include: 1) Full Hotswap/HotSpare support. RAIDZONE includes SMARTCANenclosures that provide complete monitoring and control of individualdisk drives. Under software control any given drive can be powereddown for removal and replacement without affecting the operationof any other drive. The SMARTCAN includes sensing circuitry to detectthe removal and the re-insertion of a drive and can thus automaticallyre-energize the slider unit. 2) Support for drive level RAID0, RAID1, RAID5 and RAID10. 3) RAIDZONE BIOS provides support for bootstrap of all supported RAIDlevels. 4) Expandability. Current RAIDZONE technology will allow you to configurea 40 disk drive Linux server. Using the 37GB IBM Ultra ATA 66 disk drivesthat's not too shabby. Current non-production hardware (that exists today)increases that limit to 80 disk drives. Other technology indevelopment will eliminate current limitations entirely. 5) Performance. Each disk drive has it's own bus-mastering data path tosystem memory. RAIDZONE only uses disk drives in their MASTER mode - NO SLAVES. Performance is limited by contentionfor PCI memory bandwidth and (at least under Linux 2.2.X) thedouble copy that takes place between the system buffer cacheand user space. Under Windows NT we have observed sustainedsequential read throughput of greater than 110Mbytes/sec. Bonnie numbersfor a 7-way RAID5 using 31GB IBM (7200 rpm) drives in a Intel 440BXmotherboard with dual 450Mhz PIIIs are 30Mbytes/sec writing and70 Mbytes/sec reading. 6) Administration: RAIDZONE includes a Java based GUI formonitoring and configuring the RAIDZONE disk sub-system.This can be used locally or remotely over the network. 7) Price. When compared to other full function SCSI based RAID solutionsRAIDZONE has a cost advantage due to the lower cost of Ultra ATA 66 disk drives versus SCSI disk drives with similar performance and capacity. Steve Cooper [EMAIL PROTECTED] Directory of Software Engineering,Consensys Computers Inc. -Original Message-From: Terry Ewing [EMAIL PROTECTED]To: [EMAIL PROTECTED] [EMAIL PROTECTED]Date: Wednesday, December 01, 1999 6:18 PMSubject: ide hardware raid Hello, I'm an administrator in a co-location facility and recently we had acustomer come in to replace a raid card. The only catch was the RAID cardwas IDE. This was on an NT box, but I was wondering:1) what IDE RAID cards are out there now. What is known about them.2) What is the status of Linux support for IDE RAID cards.Thanks!
Help on root fs using raid-0
I'm stuck! I've got two 1gig disks that I want to concatenate together using raid-0 and mount as / with Slackware 7 installed.I've created a minimal installation on a 60meg partition and used this to create a full installation on a raid-0 partition - /dev/md0, which I can mount and umount on /mnt to my heart's content (but which isn't much use there)!For 64,000 dollars - how can I mount /dev/md0 as / ?I've read the docs, but find them quite opaque. Surely there's an simple way! Regards: Jim Ford
Re: scsi channels (Was: ide hardware raid)
On Wed, Dec 01, 1999 at 04:20:49PM -0800, Alvin Oga wrote: they have scsi-2 raid cards from adaptec... 1-channel($430) and 3-channel($650) raid controllers.. a dumb question... what are the channels used for ??? scsi channels are different scsi busses, you use them for load balancing. btw, i doubt any of these card will work with linux :( L. -- Luca Berra -- [EMAIL PROTECTED] Communications Media Services S.r.l.
problems with RAID fs
I am using software RAID-5 on 3 IBM SCSI-UW drives and am experiencing a problem. I have moved all files from /www to /usr/raid/www, but diff is reporting many of the files to be changed. All the file sizes are correct, but it appears that often a character is swapped for another in the copy. Here are two examples: Original: td valign=top colspan=2 Copy: td valign=top cclspan=2 Original: div align="left"Tent camp facilities at high-tech comfort level /div Copy: div align="left"Tent camp facilities at high-tech cgmfort level +div This is repeated in many of the files (approximately 10%) and is stopping me from finalizing the transfer of the data. What can I do to isolate and solve this problem? Is it possible that I can get this error by screwing up the makefs I did on /dev/md0? Here are some specifics: Linux sitename 2.2.13 #2 Tue Nov 30 16:39:31 PST 1999 i686 unknown The kernel patch I applied was raid0145-19990824-2.2.11 /sbin/raidstart --version /sbin/raidstart v0.3d compiled for md raidtools-0.90 If anyone has any hints or has seen this before I'd appreciate any tips or insight you could lend. -Terry
Re: problems with RAID fs
One important thing I forgot. I copied these files by using cp -R on a big tree. Afterwards I did a diff and saw quite a few files were not exactly the same as their original. I removed the new tree and recopied. This time around I did not get the same listing from diff and the corruption was to different files. I tried copying the original tree to a dfferent, non-raid, filesystem and all files were the same as their original. I also manually used cp to copy about 10 or 12 of the corrupted files from the original tree to the RAID filesystem. After this, the files that I copied did not differ from the originals. It seems that files become corrupted under a heavy load either by the RAID5 daemon or in hardware. At 02:55 PM 12/2/99 -0800, you wrote: I am using software RAID-5 on 3 IBM SCSI-UW drives and am experiencing a problem. I have moved all files from /www to /usr/raid/www, but diff is reporting many of the files to be changed. All the file sizes are correct, but it appears that often a character is swapped for another in the copy. Here are two examples: Original: td valign=top colspan=2 Copy: td valign=top cclspan=2 Original: div align="left"Tent camp facilities at high-tech comfort level /div Copy: div align="left"Tent camp facilities at high-tech cgmfort level +div This is repeated in many of the files (approximately 10%) and is stopping me from finalizing the transfer of the data. What can I do to isolate and solve this problem? Is it possible that I can get this error by screwing up the makefs I did on /dev/md0? Here are some specifics: Linux sitename 2.2.13 #2 Tue Nov 30 16:39:31 PST 1999 i686 unknown The kernel patch I applied was raid0145-19990824-2.2.11 /sbin/raidstart --version /sbin/raidstart v0.3d compiled for md raidtools-0.90 If anyone has any hints or has seen this before I'd appreciate any tips or insight you could lend. -Terry
Re: Help on root fs using raid-0
I've done this on raid-1. I assume the procedure for raid-0 will be the same. Here is how I do it. I'm not guaranteeing this is the best way: First I'll make some assumptions. First you would like to boot to your root md0 filesystem. Second, you have a kernel that is new enough to support booting on raid and the raid-0 personality is compiled into the Kernel. Third, your copy of lilo is new enough to support booting to md devices. And 4th, you're using IDE drives. 1. Using fdisk make the partition id's of all raid partitions (on both disks) "fd". Use the "t" option in fdisk to do this. 2. Edit /etc/lilo.conf so that root=/dev/md0 and boot points to a valid boot drive. (/dev/hda in my case.) 3. Drop to your command prompt and run lilo. (Ex. Type "lilo" followed by enter at the prompt.) 4. Reboot. Be sure to have a boot disk handy. If this fails for any reason you'll need it to boot your computer. Additions or comments on this nano-procedure are welcome. Good Luck. - Original Message - From: Jim Ford To: linux-raid Sent: Thursday, December 02, 1999 10:59 AM Subject: Help on root fs using raid-0 I'm stuck! I've got two 1gig disks that I want to concatenate together using raid-0 and mount as / with Slackware 7 installed. I've created a minimal installation on a 60meg partition and used this to create a full installation on a raid-0 partition - /dev/md0, which I can mount and umount on /mnt to my heart's content (but which isn't much use there)! For 64,000 dollars - how can I mount /dev/md0 as / ? I've read the docs, but find them quite opaque. Surely there's an simple way! Regards: Jim Ford
