Re: ide and hot swap
Marc Merlin wrote: While I'll be the last person to praise IDE, recent drives and controllers have CRC error checking, which is actually better than parity. OK, maybe this could shed some light on a mystery I'm fighting here with. I recently replaced this (Asus P55T2P4) motherboard: --- Oct 22 15:53:45 ruri kernel: PIIX3: IDE controller on PCI bus 00 dev 39 Oct 22 15:53:45 ruri kernel: PIIX3: not 100% native mode: will probe irqs later Oct 22 15:53:45 ruri kernel: ide0: BM-DMA at 0xe800-0xe807, BIOS settings: hda:pio, hdb:pio Oct 22 15:53:45 ruri kernel: ide1: BM-DMA at 0xe808-0xe80f, BIOS settings: hdc:pio, hdd:pio Oct 22 15:53:45 ruri kernel: hda: IBM-DTTA-351680, ATA DISK drive Oct 22 15:53:45 ruri kernel: hdc: IBM-DTTA-351680, ATA DISK drive Oct 22 15:53:45 ruri kernel: ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 Oct 22 15:53:45 ruri kernel: ide1 at 0x170-0x177,0x376 on irq 15 Oct 22 15:53:45 ruri kernel: hda: IBM-DTTA-351680, 16124MB w/462kB Cache, CHS=32760/16/63, (U)DMA Oct 22 15:53:45 ruri kernel: hdc: IBM-DTTA-351680, 16124MB w/462kB Cache, CHS=32760/16/63, (U)DMA --- It ran flawlessly and the disks showed no problems running with DMA and delivering around 10MB/s or more with bonnie. Now I replaced the motherboard (and nothing else like cables) with a Tyan Trinity 100AT one: --- Nov 7 18:25:51 ruri kernel: VP_IDE: IDE controller on PCI bus 00 dev 39 Nov 7 18:25:51 ruri kernel: VP_IDE: not 100% native mode: will probe irqs later Nov 7 18:25:51 ruri kernel: ide0: BM-DMA at 0xe000-0xe007, BIOS settings: hda:DMA, hdb:DMA Nov 7 18:25:51 ruri kernel: ide1: BM-DMA at 0xe008-0xe00f, BIOS settings: hdc:DMA, hdd:DMA Nov 7 18:25:51 ruri kernel: hda: IBM-DTTA-351680, ATA DISK drive Nov 7 18:25:51 ruri kernel: hdc: IBM-DTTA-351680, ATA DISK drive Nov 7 18:25:51 ruri kernel: ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 Nov 7 18:25:51 ruri kernel: ide1 at 0x170-0x177,0x376 on irq 15 Nov 7 18:25:51 ruri kernel: hda: IBM-DTTA-351680, 16124MB w/462kB Cache, CHS=32760/16/63, UDMA Nov 7 18:25:51 ruri kernel: hdc: IBM-DTTA-351680, 16124MB w/462kB Cache, CHS=32760/16/63, UDMA --- And promptly got this when putting some load on the HDs: --- Nov 7 17:54:00 ruri kernel: hdc: dma_intr: status=0x51 { DriveReady SeekComplete Error } Nov 7 17:54:00 ruri kernel: hdc: dma_intr: error=0x84 { DriveStatusError BadCRC } Nov 7 17:54:00 ruri kernel: hda: dma_intr: status=0x51 { DriveReady SeekComplete Error } Nov 7 17:54:00 ruri kernel: hda: dma_intr: error=0x84 { DriveStatusError BadCRC } --- Turning off UDMA solves the problem and also halves the performance. My questions are thus: Did the old chip-set just use DMA (instead of UDMA) and thus didn't triggers these problems? If so, any way to put the VIA controller into plain DMA mode? Or did the old chip-set no CRC checks and thus never noticed these problems? But that should have caused some corruption and it never occured. Dewa, CB -- // CB aka Christian Balzer, Tannenstr. 23c, D-64342 Seeheim, Germany \X/ [EMAIL PROTECTED] | Voice: +49 6257 83036, Fax/Data: +49 6257 83037 SWB - The Software Brewery - | http://www.swb.de/ | Anime no Otaku
Re: ide and hot swap
Gerrish, Robert [EMAIL PROTECTED] wrote: backups can save your business. You not only should think about system failures, but fires, floods, etc. and silent data corruption. Remember, IDE hasn't parity. Also, disk-drives can develop bad blocks over time. That's why incremental backups, even done on your old QIC-120 for only the most important files, can be good.
Re: ide and hot swap
On mar, nov 09, 1999 at 10:42:53 +0100, Marc SCHAEFER wrote: and silent data corruption. Remember, IDE hasn't parity. Also, disk-drives can develop bad blocks over time. While I'll be the last person to praise IDE, recent drives and controllers have CRC error checking, which is actually better than parity. Marc -- Microsoft is to software what McDonalds is to gourmet cooking Home page: http://marc.merlins.org/ (friendly to non IE browsers) Finger [EMAIL PROTECTED] for PGP key and other contact information
Re: ide and hot swap
While I'll be the last person to praise IDE, recent drives and controllers have CRC error checking, which is actually better than parity. would you happen to know which drives and controllers? The promise udma66's? Any WD IDE's or IBM's 36+gig. -sv
RE: ide and hot swap
Seth Vidal wrote: We've got DLT's doing backups right now and we're conceiving that it might be cheaper to setup a system with 2 or 3 linear striped or raid 0 34+gig ide disks and have 2 sets of these disks that we swap out week to week for backups - rather than spend a fortune in DLT tapes and deal with a whopping 4MB/s transfer time. We would be using a set of disks for 4 weeks then swapping out to another set - the other set would be fresh formatted at that point and would be ready to go for the next month's backups. Price wise, this seems like a good approach. If it were my system, I would be concerned about disaster recovery. I have been a believer for a long time in tape rotation and offsite storage. Also, you are risking losing 4 weeks worth of data; a full backup at least weekly and incremental backups can save your business. You not only should think about system failures, but fires, floods, etc. An onsite disk storage scheme doesn't take these situations into account. Perhaps, if you want to consider alternate storage, you should look at optical media or some other approach. Bob Gerrish
RE: ide and hot swap
Price wise, this seems like a good approach. If it were my system, I would be concerned about disaster recovery. I have been a believer for a long time in tape rotation and offsite storage. Also, you are risking losing 4 weeks worth of data; a full backup at least weekly and incremental backups can save your business. You not only should think about system failures, but fires, floods, etc. An onsite disk storage scheme doesn't take these situations into account. Perhaps, if you want to consider alternate storage, you should look at optical media or some other approach. but I'm talking about doing full rotations. Right now we're loading 7 tapes into the DLT jukebox and it rotates for a month through those. Then the level 0 tapes come out and go to my house for offsite storage. We start over (more or less) every month. I was proposing using 2 or 3 big ide's every month. We do risk losing 4 weeks from a fire but we Always have risked that. But all other storages are offsite. You risk losing whatever's in the room at the time of a fire. While 4 weeks is greater than 1 day it might be a tolerable risk. my biggest concern is MTBF and not fires. Fires are a mess but your data (while important) is not the first concern after a fire - rebuilding is. Its the random drive failures and overwrites that I think most backups protect from. -sv
ide and hot swap
I know this has been covered before but I can't find a searchable archive of this list anywhere. We've got DLT's doing backups right now and we're conceiving that it might be cheaper to setup a system with 2 or 3 linear striped or raid 0 34+gig ide disks and have 2 sets of these disks that we swap out week to week for backups - rather than spend a fortune in DLT tapes and deal with a whopping 4MB/s transfer time. We would be using a set of disks for 4 weeks then swapping out to another set - the other set would be fresh formatted at that point and would be ready to go for the next month's backups. What I'm most interested in is if anyone has seen. 1. external inclosures for ide disks and if there is any hotswap support for ide devices in the kernel. I figure 4 disks across 2 controllers while not an optimum situation would see a HUGE speed boost over DLT drives. And with IDE drive prices dropping like crazy We'd eventually be at a point where the media would almost be cheaper than DLT media. With potential expansion of ide drives into the 100's of gigabyte range this might make a worthwhile effort. any ideas/suggestions. -sv
Hot swap problems... continued
Hi again, well today I repeated the test pulling out the disk and waiting a longer time - with no success. I did a raidhotremove on the failed partitions and then added them back with raidhotadd, a cat of /proc/mdstat revealed a quickly increasing finish time to recovery of the mirror md3 : active raid1 sdb8[2] sda8[0] 6707008 blocks [2/1] [U_] recovery=0% finish=119606.5min (and rising) at this stage I tried to unmount the partition (/home) and this hung - not good. Luckily I have the magic sysreq installed so I was able to sync, unmount and reboot cleanly. Any idea what could be happening? Brian Murphy
Re: Hot swap problems... continued
Brian Murphy wrote: md3 : active raid1 sdb8[2] sda8[0] 6707008 blocks [2/1] [U_] recovery=0% finish=119606.5min (and rising) I have also observed this problem. It seem to be specifically related to a RAID1 configuration going into degraded mode with one disk in array. It certainly seem to be a bug. I haven't used this configuration with latest raid patch so I don't know if it's been fixed yet. tomas/
Re: Hot swap problems... continued
On Fri, 24 Sep 1999, Brian Murphy wrote: Hi again, well today I repeated the test pulling out the disk and waiting a longer time - with no success. I did a raidhotremove on the failed partitions and then added them back with raidhotadd, a cat of /proc/mdstat revealed a quickly increasing finish time to recovery of the mirror md3 : active raid1 sdb8[2] sda8[0] 6707008 blocks [2/1] [U_] recovery=0% finish=119606.5min (and rising) I have fiddled with this hotswapping thing also. My experience was that after removing adding the performance degraded to like 1/10th of the original. I figured out that it might have to do something about scsi defaulting back to async transfer mode after adding. Also noticed that when i really changed the physical disk to the scsi bus, few times everything went find and i was able to start recovery process but like 40% of the cases my kernel panicked when the raid accessed first time the new replaced disk. Tho the bug might still sit between keyboard and chair. So what i am asking is that what is the _correct_ procedure to add a scsi device on the fly to to the bus? I did something like this: (if i remember correctly) raidsetfaulty /dev/md0 /dev/sda1 echo "scsi remove-single-device 0 0 0 0" /proc/scsi/scsi this point i made the physical change operation of disk echo "scsi add-single-device 0 0 0 0" /proc/scsi/scsi at this point the kernel correctly did found the addded device raidhotadd /dev/md0 /dev/sda1 sometimes got panic in here Few times the recovery process started, few times the kernel just paniced. Can someone straight ahead say what i am doing wrong ? -- Mika
Re: Hot swap problems... continued
Mika Kuoppala wrote: raidsetfaulty /dev/md0 /dev/sda1 Shouldn't there be a raidhotremove /dev/md0 /dev/sda1 before low level removal of the scsi device? Could explain sudden panics, if an unexpected state occurs. tomas/
Re: Hot swap problems... continued
At 05:02 PM 9/24/1999 +0200, you wrote: Mika Kuoppala wrote: raidsetfaulty /dev/md0 /dev/sda1 Shouldn't there be a raidhotremove /dev/md0 /dev/sda1 before low level removal of the scsi device? Could explain sudden panics, if an unexpected state occurs. That would make sense === David Cooley N5XMT Internet: [EMAIL PROTECTED] Packet: N5XMT@KQ4LO.#INT.NC.USA.NA T.A.P.R. Member #7068 We are Borg... Prepare to be assimilated! ===
Re: Hot swap problems... continued
Stephen Waters wrote: how about a nice hotswap step-by-step in the howto... Sure. Does that mean you're a volunteer? 8) tomas/
Re: Hot swap problems... continued
i'm an end-user who is curious about how to do this since i am in the process of buying a 4 18GB drive scsi system for use w/ software raid5. my main problems are related to the scsi commands to issue and what order they go in. if they're in a howto, then when i get the drives in i can check out how well it works. -stephen Tomas Fasth wrote: Stephen Waters wrote: how about a nice hotswap step-by-step in the howto... Sure. Does that mean you're a volunteer? 8) tomas/
HOT-SWAP
Thanks, the hot-swap on IDE disks really works. I disconnected one of the IDE cables to create a failed disk. After that I wanted to get the disk back in the array without a reboot of the system. One of you gave me the advice to do a "raidhotremove" first and after that the "raidhotadd". And yes it worked :-) The reconstruction process started without any problems. - Alex
Re: Hot Swap
Helge Hafting wrote: [question on how to hot-swap IDE] I have an IDE drive that I can disconnect and reconnect with no problems. It is not on a RAID though. The controller is nothing special, a ultra-dma thing built into the shuttle motherboard. The trick seems to be unloading the IDE kernel module before reconnecting. It probes the drive just fine upon reloading. Unfortunately you can't do that with a IDE root file system. Helge Hafting I did some hot-swap tests with IDE drives successfully: normally the IDE driver is compiled fix in the kernel , and even if you compile it as a module in the 2.2.x kernels , you can not unload it because its busy in the case the root-fs is on the IDE disk. The kernel probes the attached disks at boot time, so it has a static table of the disks which are present in your machine. I tested the following: a soft-RAID5 array consisting of 3 disks , hda1 , hdb1 , hdc1 diring heavy writes on the raid array, I disconnected the power hdb for example. At this point the kernel sees that the device does not respond and writes many messages in the syslog. The raid layer detects the error condition , marks the device as faulty (you can see this in /proc/mdstat ) , and continues in degrated mode. at this point you must issue a: raidhotremove /dev/md0 /dev/hdb1 which removes the hdb1 device from the array (again, check your /proc/mdstat ) now you can reconnect the disk if it's a new blank disk you have to repartition the disk like the old one , creating hdb1 for example, and you can do this on the fly without a reboot (it worked for me, I know that there are disks where you must reboot to ensure that the new partition table is updated properly) now simply type: raidhotadd /dev/md0 /dev/hdb1 et voila': the device is added to the array and hot-reconstruction begins in background and in /proc/mdstat you can see the progress status of the reconstruction. regards, Benno.
Re: Hot Swap
not sure if it's a good idea to do this with any disk interface that not's designed for it. i have a DEC paper somewhere about SCSI, and it spends a lot of time on hot swap issues. Hot swap can be extremely bad for an interface physically and electrically. eg, SCA and other proprietary scsi hot-swap interconnects are designed specifically with hot swap in mind, and even still (according to the DEC paper) there are non-insignificant risks to hot swapping a SCA device. Most of which relate to the timing of the (de)coupling of power lines relative to ground lines. (IIRC). this is why there are actually 2 different types of sca connector, the newer one is compatable with the older, and the only real difference is the drive and data ground pins are contacted first, then the drive power 12v and Vcc, then the data Vcc. hard to tell the diff by looking, though. allan
Hot Swap
Hello RAID group, I have a Software RAID-1 system here with two IDE drives. The md1 device contains a complete mirror of a RedHat 5.2 root file system. I boot the system with a bootdisk created while md1 was mounted /. The kernel I use is 2.2.3. When the system is running i disconnect one of the IDE cables {';'} oops... Now when I put the IDE cable back on its place I try to raidhotadd the "failed" disk. And I get a: device is busy error. After a reboot of the system its no problem to get the disk back in the array. Is there a way of skipping the "reboot of the system" part ? - Alex
Re: Hot Swap
I doubt it, unless the controller is hotswap capable and you can reload the IDE driver I don't know of any hotswap capable IDE controllers (Not to say that I wouldn't be interested if anyone else on the list does!!!) On Thu, 22 Apr 1999, a.loots wrote: Hello RAID group, I have a Software RAID-1 system here with two IDE drives. The md1 device contains a complete mirror of a RedHat 5.2 root file system. I boot the system with a bootdisk created while md1 was mounted /. The kernel I use is 2.2.3. When the system is running i disconnect one of the IDE cables {';'} oops... Now when I put the IDE cable back on its place I try to raidhotadd the "failed" disk. And I get a: device is busy error. After a reboot of the system its no problem to get the disk back in the array. Is there a way of skipping the "reboot of the system" part ? - Alex James ([EMAIL PROTECTED]) It doesn't run on an open source platform, therefore it, by definition, does not matter.
Re: Hot Swap
I doubt it, unless the controller is hotswap capable and you can reload the IDE driver I don't know of any hotswap capable IDE controllers (Not to say that I wouldn't be interested if anyone else on the list does!!!) I would as well be interested in a hot-swappable ide controller. -sv
Re: hot swap drive carriers
Just wondering what the linux raid folk are using for hot swappable carriers. I haven't been able to find a brand I'm happy with (tried Antec and Kingston brands). k Try megadrive (www.megadrive.com). we have some for a proprietary Panasonic nonlinear editor, and they are of very good quality, hot swap capable, ultra 2 and fiberchannel available.. dave
hot swap drive carriers
Just wondering what the linux raid folk are using for hot swappable carriers. I haven't been able to find a brand I'm happy with (tried Antec and Kingston brands). Chris -- Crack the RSA RC5-64 encryption algorithm. Use your idle cpu cycles to take part in the fastest computer on the planet. See: http://www.distributed.net Join: http://www.execpc.com/~silby/teamamd.html