Re: FAQ
On Fri, Aug 04, 2000 at 09:48:18AM +0530, Abhishek Khaitan wrote: > Can;t we use bunzip2 instead of playing with tar? And after bunzip2, try tar > -x kernel-2.2.16.tar ? The usual suggestion is: bzip2 -dc | tar -xf - s/bzip2/gzip/ or s/bzip2/uncompress/ as necessary -- Randomly Generated Tagline: If you remove stricture from a large Perl program currently, you're just installing delayed bugs, whereas with this feature, you're installing an instant bug that's easily fixed. Whoopee. -- Larry Wall in <[EMAIL PROTECTED]>
Re: Problem with raid and new kernel
On Tue, Jul 18, 2000 at 03:20:58PM +0300, Dimitrios Stergiou wrote: > I compiled kernel 2.2.16, included all RAID[0,1,2,3,4] as modules, cretaed > an initrd image (mkinitrd /boot/initrd-2.2.16-10.img 2.2.16-10) and > rebooted. > > The system, under no circumstances booted. It complained about "md0 not > started" and "unable to mount root fs" When you made the initrd, did you do "--with=raid0"? (and any other raid levels you need for the root drive)? -- Randomly Generated Tagline: Your mother seems really upset. I better go have a talk with her -- during the commercial. -- Homer Simpson Simpsoncalifragilisticexpiala(annoyed grunt)ocious
Re: Abit KA7 + RAID
On Fri, Jul 14, 2000 at 05:24:27PM -0400, Edward Schernau wrote: > I saw a blurb somewhere about this board offering built in > RAID 0 and 1, a BIOS thing. Is this just more WinRAID, like > the Promise Fasttrak? I have a KA7-100 and there is nothing RAID-related in the BIOS. -- Randomly Generated Tagline: Let him who is without shit cast the first turd. -- Larry Wall in <[EMAIL PROTECTED]>
Re: About RAID
On Tue, Jun 27, 2000 at 09:40:51AM +0200, Chencho wrote: > How can I add a disk to a raid 5.I have created a raid 5, > but I dont know how add a new disk. Assuming you mean "add a new disk to expand the size of the array", at the moment: 1) backup the data 2) remove and re-create the RAID array w/ the new disk 3) restore the data -- Randomly Generated Tagline: "I don't even have to get dressed up for Halloween. I go as me." - Judge Judy
Re: Patches for 2.2.16?
On Fri, Jun 09, 2000 at 05:51:43AM -0400, Mike Black wrote: > I've been running 2.2.16 with the 2.2.15-A0 patch for about 18 hours now on > two boxes (one RAID1, one large RAID5). > It's working fine (the md.c rejects don't matter -- that was for old version > of md.c). Well, I tried the "2.2.15, raid-2.2.15-A0 patch, 2.2.16 patch, build" method. Everything compiled, so I assume everything's good. there was the single patch failure against md.c when going 2.2.15 to 2.2.16, but hopefully that's ok. The box is remote, so I'm waiting until a little later tonight to go ahead and reboot it, just in case it doesn't come back up for one reason or another. -- Randomly Generated Tagline: "When you say 'I wrote a program that crashed Windows,' people just stare at you blankly and say 'Hey, I got those with the system, *for free*.'" - Linus Torvalds
Patches for 2.2.16?
I'm about to install the 2.2.16 kernel to fix the capabilities bug, and found that the 2.2.15-A0 raid patch fails in 2 places: patching file `include/linux/sysctl.h' Hunk #1 FAILED at 429. 1 out of 1 hunk FAILED -- saving rejects to include/linux/sysctl.h.rej patching file `drivers/block/md.c' Hunk #2 FAILED at 22. Hunk #3 succeeded at 3034 (offset 2 lines). Hunk #5 succeeded at 3833 (offset 2 lines). Hunk #7 succeeded at 3888 (offset 2 lines). 1 out of 7 hunks FAILED -- saving rejects to drivers/block/md.c.rej patching file `arch/sparc64/kernel/sparc64_ksyms.c' Reversed (or previously applied) patch detected! Assume -R? [n] Apply anyway? [n] Skipping patch. 2 out of 2 hunks ignored -- saving rejects to arch/sparc64/kernel/sparc64_ksyms.c.rej the sysctl one seems easy to fix, but the drivers/block/md.c one is rather substantial. since I'm running on a pentium, I don't care about the sparc64 rejects. I wasn't sure if there was a 2.2.16 patch coming out soon, and I wasn't sure I wanted to install a "A0" patch ... Any thoughts? -- Randomly Generated Tagline: "Of course my password is the same as my pet's name. My cat's name was Q47pY!3, but I change it every 90 days." - Roddy Vagg
Re: RAID 1+0
On Thu, Jun 01, 2000 at 10:23:21AM +0100, Corin Hartland-Swann wrote: > So, is 0+1 the only combination currently allowed? To my knowledge, yes. > Is anybody else interested in seeing 1+0, 5+0, etc? Personally, I would say that if you're going to go for 5+0 or 5+1, you should really get HW RAID 5 and then use the SW RAID to either stripe or mirror that. I think since 0+1 is supported, 1+0 should be supported as well. -- Randomly Generated Tagline: But we can both blame it all on Henry. -- Larry Wall on perl's regex engine
Re: RAID 1+0
On Wed, May 31, 2000 at 10:17:16AM -0400, Theo Van Dinter wrote: > NxP: 1/(PN-1) vs N/(PN-1) Just to correct myself -- this equation actually doesn't work after thinking about it. It works for P=2, but after that, the whole game changes... Regardless, striped mirrors is usually considered better than mirrored stripes. 8) -- Randomly Generated Tagline: "I don't like rap because I'm stuffy and british." - James Burke
Re: RAID 1+0
On Wed, May 31, 2000 at 09:10:30AM -0400, Andy Poling wrote: > That's the error you will get any time that you try to layer raid levels > that md does not support layering. It's a safety belt mechanism of sorts. Arguably, any combination should be allowed, but 0+1 and 1+0 at minimum. > Either way, specific pairs of disks can fail without consequence. With 0+1, > sda & sdb can fail or sdc & sdd can fail. > > 0+1 and 1+0 are equally safe as far as that's concerned... i.e. not very. :-) There are two major benefits of striped mirrors over mirrored stripes (I forget which is 0+1 and which is 1+0, so I'll just be verbose): 1) reconstruction is much faster (resync the failed disk, not the failed array). 2) the chance that the whole array will fail is less, even though certain 2 disk failures will take either array down. Taking the 2x2 example: With either striped mirrors or mirrored stripes, the chance that the first disk will take down the array is 0% (the mirror will cover it). However, with striped mirrors, there's only a 1 in 3 chance that the next disk failure will take the array down (both disks in the same mirror have to die). Mirrored stripes, though, gives you a 2 in 3 change that the second disk failure takes the array down. (the first disk knocked out the first stripe, so any failure in the second stripe brings the whole thing down.) These values get more disjointed as the arrays grow: 3x2: 1/5 vs 3/5 4x2: 1/7 vs 4/7 ... Nx2: 1/(2N-1) vs N/(2N-1) NxP: 1/(PN-1) vs N/(PN-1) Striped mirrors is costly for administration (mirrored stripes is always 3 arrays (mir,mir,stripe), whereas striped mirrors depend on the number of disks (Nx2 is N+1 arrays, n mir + stripe)), but the costs are worth it IMHO. -- Randomly Generated Tagline: "Perl is your friend. Use Perl." - Randal Schwartz
Re: I need your help
On Thu, Apr 27, 2000 at 05:58:19PM +0200, Andreas Martmann wrote: > Here is my problem: I have destroyed my Bootblock. Now I can´t access > the root-Partition in order to make a new one. The only thing that I can > reach is the kernel i had put in the ext2-Partition. If you can get the kernel, you should be able to make a boot disk with that kernel that then boots the system up normally (assuming you don't need kernel modules for bootup -- even then, you can probably use the new bootdisk to boot the rescue disk...) Alternatively, you could download a boot disk from RedHat which has the RAID drivers built in and see if you can get something up and running from there. -- Randomly Generated Tagline: "My mother never saw the irony in calling me a son-of-a-bitch." - Richard Jeni
Re: fsck'ing RAID's
On Mon, Apr 24, 2000 at 10:24:20PM +0200, Jakob Østergaard wrote: > Resync shouldn't change what is read from the array, as it only rebuilds the > parity -- the redunant information -- and doesn't affect the ``real'' data. It depends on which RAID level and which disk fail. In this case (RAID5), you're going to have to rebuild both parity *AND* data (this isn't RAID [2-4]...) While the fsck should see the same information whether or not a resync needs to occur, it's going to be *much* slower to fsck during a resync than after the resync is completed. (not to mention that the fsck will have to recreate the data to check -- hopefully the rebuild process will use this so it doesn't have to recreate the data twice.) -- Randomly Generated Tagline: "Capital punishment turns the state into a murderer. But imprisonment turns the state into a gay dungeon-master." - Emo Philips
Re: Chunk size in mirrored configurations?
On Fri, Apr 14, 2000 at 12:06:47PM -0700, Erich wrote: > I looked through the documentation, and I can't find any good > information about what the chunk size should be in a mirrored > configruation. I'm using three disks in a Level 1 configuration. The well, from the man page: chunk-size size Sets the stripe size to size bytes. Has to be a power of 2 and has a compilation-time maximum of 4M. (MAX_CHUNK_SIZE in the kernel driver) typical values are anything from 4k to 128k, the best value should be determined by experimenting on a given array, alot depends on the SCSI and disk configura tion. Since a mirror isn't striped, I'd say it doesn't matter. I personally used 32 in my setup, but ... -- Randomly Generated Tagline: "I thought you were dead. Yeah ... I get that a lot." - From the movie "Alien: Resurrection"
Re: The meaning of this?
On Thu, Apr 13, 2000 at 06:28:04PM +0200, [EMAIL PROTECTED] wrote: > > I think it means you tried to run fsck across an area being resynced. > > Is this bad? Should the init scirpts be modified to somehow avoid this? If doing this is causing problems, there's a bug in the RAID code. At the filesystem layer, the "device" that the filesystem is on should either be there, or it shoudn't be there. A resync should be transparent to everything else (except that there's a bunch of disk i/o ...) -- Randomly Generated Tagline: "It's kind of like wanting to be in a band, but being a roadie ..." - Instructor Otten
Re: Adding a spare-disk to a RAID5 array?
On Tue, Apr 04, 2000 at 10:28:47PM +0100, Darren Nickerson wrote: > I've found some cash, and want to add a spare disk to our raid5 array for > added redundancy. > > Can this be done? It is a matter of > > 1. raidstop > 2. add spare to raidtab > 3. raidhotadd spare To add a spare? I'm fairly certain (assuming the disk/partition is the same size as the rest of the stripe) it's just: 1. add disk to system 2. raidhotadd spare If you aren't using autostarting arrays, then you'll probably need to add the spare-disk to raidtab. -- Randomly Generated Tagline: "You're basically killing each other to see who's got the better imaginary friend." - Richard Jeni (on going to war over religion)
RAID Devices and FS labels
On my home machine today, I decided to change how the filesystems are listed in /etc/fstab from the standard /dev/name to FS labels: LABEL=ROOT / ext2defaults1 1 LABEL=USR /usrext2defaults1 2 I did this mostly because I started to move partitions around a bit this morning. The labels, of course, work fine on my non-RAID home setup, but I started pondering what would happen on my mirrored server disks since there would be 3 devices with the same FS label (md#, sda#, and sdb# in my case) for each mount point. Does anyone know how the tools would handle this situation? I'd assume that given a list of devices and labels, the RAID devices would come up first, and then the individual partitions, but I'm not sure how this works WRT mount, fsck, etc. Any ideas? Thanks. -- Randomly Generated Tagline: Cyberspace: The next best thing to being there.
Re: Disk v. Tape Backup -- Re: root on RAID
On Fri, Mar 31, 2000 at 02:56:57PM -0800, Gregory Leblanc wrote: > un-planned need to restore an entire system from tape. Usually the restores > that I do are because Joe User deleted his all important spreadsheet, and > NEEDS to have it back. I definately agree that RAID shouldn't (and can't) > replace tapes. I use RAID to protect my systems, and tapes to protect the > data that's housed on those systems. I don't consider my OS as data, but > the config files are definately data (overwriting smb.conf is not good...). So what you're saying is that you want a logged fs w/ snapshots on a RAID array and off-site tape backups... Everything you want in a convenient package: on-line "backups" for quick recovers of recent data (up to the last snapshot), long-term backups off-site for DR, and resilience against disk failures through RAID. -- Randomly Generated Tagline: For the sake of argument I'll ignore all your fighting words. -- Larry Wall in <[EMAIL PROTECTED]>
Re: root on RAID
On Fri, Mar 31, 2000 at 11:42:01AM -0800, Gregory Leblanc wrote: > Mostly because it's (pardon my French) a bitch to recover from. RAID5 and ??? > (although still not as easy as a plain mirror), while a stripe of two > mirrors (RAID01) is a real pain to recover from. Mostly this applies to ??? How are you defining "recover from"? For a single disk failure, you're fine no matter what you do (unless you go linear/RAID0). Plug a replacement disk in, and you're hoopy. Are you talking about going from RAID to non-RAID? -- Randomly Generated Tagline: "We all know Linux is great...it does infinite loops in 5 seconds." (Linus Torvalds about the superiority of Linux on the Amterdam Linux Symposium)
Re: root on RAID
On Fri, Mar 31, 2000 at 09:27:59AM +, Glenn Hudson wrote: > The installation of Red Hat Linux's root partition onto a RAID device is > not supported. > > Do you know what problems having the root partition on RAID will cause? There are no problems, but the Redhat installer doesn't handle it. (the 6.2 one does I believe). I have 6.1 running on root RAID, there aren't any problems. (I had to install to a non-RAID disk, then create a mirror w/ a failed disk, copy the data over, boot off the mirror, and then hot-add the single disk back in.) If you want RAID5 or something for the root disk, you have to seperate /boot (or where your kernel is located) first since LILO can't deal with RAID directly. -- Randomly Generated Tagline: "The random quantum fluctuations of my brain are historical accidents that happen to have decided that the concepts of dynamic scoping and lexical scoping are orthogonal and should remain that way."- Larry Wall
Re: Raid5 with two failed disks?
On Thu, Mar 30, 2000 at 02:21:45PM -0600, Bill Carlson wrote: > 1+5 would still fail on 2 drives if those 2 drives where both from the > same RAID 1 set. The wasted space becomes more than N/2, but it might > worth it for the HA aspect. RAID 6 looks cleaner, but that would require > someone to write an implementation, whereas you could do RAID 15 (51?) > now. 2 drives failing in either RAID 1+5 or 5+1 results in a still available array: minimal RAID 1+5 (ie: mirroring stripes) stripe 1mirroredstripe 2 sda1sdd1 sdb1sde1 sdc1sdf1 If 2 in the same stripe die, then that stripe dies, but the array is still there since the other stripe is fine. If 1 in each stripe die then both stripes are still available (although degraded) and the whole array is still up. You can lose a third disk as well without any problems (either all 3 on one side, or a 1/2 split which leaves the array available but degraded.) minimal RAID 5+1 (ie: striped mirrors) sda1mirroredsdd1 sdb1mirroredsde1 sdc1mirroredsdf1 and then striped vertically. 2 disks failing in the same mirror means the array goes into degraded mode, but it's still available. 2 disks failing in different mirrors means that the array is still 100% up and available (not degraded). you can still lose a third disk, either both sides of a mirror and another, or 1 from each mirror -- same result, the array is still available (and in the latter case, non degraded). In either case, losing a 4th disk could potentially bring the array down. I'll agree, BTW, that this is a large amount of "wasted" space, but it depends what your goals are. x2 disks may be worth it if you need a large amount of reliability. I haven't looked at them too much, but since RAID 6/7(?) handle double disk failures, they're worth looking into. > My thought here is leading to a distributed file system that is server > independent, it seems something like that would solve a lot of problems It would be pretty nifty. -- Randomly Generated Tagline: "If you're ordering from us, your miserable enough to do without spam from strangers." - Despair.com's privacy statement
Re: Raid5 with two failed disks?
On Thu, Mar 30, 2000 at 08:36:52AM -0600, Bill Carlson wrote: > I've been thinking about this for a different project, how bad would it be > to setup RAID 5 to allow for 2 (or more) failures in an array? Or is this > handled under a different class of RAID (ignoring things like RAID 5 over > mirrored disks and such). You just can't do that with RAID5. I seem to remember that there's a RAID 6 or 7 that handles 2 disk failures (multiple parity devices or something like that.) You can optionally do RAID 5+1 where you mirror partitions and then stripe across them ala RAID 0+1. You'd have to lose 4 disks minimally before the array goes offline. -- Randomly Generated Tagline: "There are more ways to reduce friction in metals then there were release dates for Windows 95."- Quantum on TLC
resizing raid arrays
Since this thread has popped up again, here's the URL I was referring to in my previous email: http://ostenfeld.dk/~jakob/Software-RAID.HOWTO/ You can resize RAID0 arrays, but so far not RAID5 arrays. 8( -- Randomly Generated Tagline: It's the Magic that counts. -- Larry Wall on Perl's apparent ugliness
Re: ext2resize
On Wed, Mar 29, 2000 at 03:49:17PM -0500, David Holl wrote: > I would 'hope' it would work. (under the assumption that raid is only > concerned with portraying a block device without concern for what is > stored on that block device) Of course, that's just a 'hope'. :) Unfortunately, with the RAID superblock at the back of each RAID partition, you're going to need a tool that understands that it's there. I remember seeing someone posting about a RAID resizer which calls the resize2fs package. Don't know where that went, but that's probably what you want. -- Randomly Generated Tagline: "... by changing many lightbulbs, and I'm an Electrical Engineer, and it only takes 1 of us ..." - Prof. Vaz
Re: Changing controllers strategy?
On Tue, Mar 28, 2000 at 12:47:08PM -0500, Seth Vidal wrote: > > I'm a bit cautious here as I've had a bad experience when experimenting > > with disk changing and ended up with a corrupted array. Is it just me, or should the RAID superblock include information to make disk ordering unimportant? BTW: worst case (Assuming RAID 4/5), you can fail one partition, move it to the new controller/new name, then add it back. repeat as necessary. -- Randomly Generated Tagline: I don't know if it's what you want, but it's what you get. :-) -- Larry Wall in <[EMAIL PROTECTED]>
Re: product testimonials
On Mon, Mar 20, 2000 at 10:36:40AM -0800, [EMAIL PROTECTED] wrote: > while true > do > sleep 3 > if [ -n "`cat $RAID_STAT | perl -ne 'if (/(.*\[U*[^\]\[U]+U*\])$/) { print >\"Failure! $1\n\"; }'`" ] > then > cat $RAID_STAT | mail -s " Raid Failure Warning " $ADMIN_EMAIL > sleep 600 > fi > done Ugh. Why would you want to check every _3 seconds_ for a disk failure? That's a lot of wasted cycles. Instead of forking a few different programs like that, how about: a cronjob that runs every 5 minutes and just diffs mdstat and a copy of a good mdstat: */5 * * * * /usr/bin/diff /mdstat.good /proc/mdstat Just do "cp /proc/mdstat /mdstat.good", and that's it. You'll get a mail if there's a difference (including failure), and no mail if they're the same. You still get mailed every 5 minutes, but you don't use quite so many cycles and processes to do it. (worst case, you have a 5-10 minute window with a failed disk. If you're worried about it, change the */5 to * ...) You could also throw this into a monitor using "mon" (http://www.kernel.org/software/mon) which you can set to mail you once an hour/etc. -- Randomly Generated Tagline: I'd love to, but I want to spend more time with my blender.
Re: New(?) IDE hardware RAID device
On Wed, Feb 09, 2000 at 11:26:20AM -0800, Gregory Leblanc wrote: > True, but even with the nifty patches that RedHat has supplied, you can only > boot from RAID 1. I was thinking you could grab two cards like this, and > create RAID-0 arrays on both, and then mirror those using Linux software > RAID 1 to give you a nice RAID 10 redundant, fast drive array. It does > sound a bit expensive though... You can boot off of non-RAID 1 arrays, you just need to finese things a little bit. For instance, make /boot a RAID-1 array and point LILO at it. You can then make a RAID-5 or 10 array for / and everything else. It should be able to boot up without any problems. -- Randomly Generated Tagline: They can always run stderr through uniq. :-) -- Larry Wall in <[EMAIL PROTECTED]>
Re: RAID 1 setup
On Tue, Feb 08, 2000 at 10:40:46AM -0500, Sean Millichamp wrote: > the RAID. I could do it, but it would be a big pain in the butt and it > seems that for RAID-1 it should be possible without formatting my original > drive. > > I looked for mention of a setup like this in the howto but didn't see any > sign of it. Has anyone done this? Is it possible? Well ... Unfortunately since the Linux SW RAID modifies the partition, you have to move data around before you can use the original disks again. (boo!) What I ended up doing was to make the RAID array using a new disk, and then specifying the original disk as "failed-disk". After the array is created, do a mke2fs on the new arrays. I would made a boot rescue disk (w/ the raid drivers), booted up to the rescue disk, and copied all of the data over to the new array. To boot, you can either use the boot floppy (linux root=/dev/md0) or setup LILO to boot to either disk. If everything is running fine on the array, you can then repartition the original disk (if neccesary), do a 'raidhotadd /dev/md# /dev/XXX' and sync the partitions together. Re-install LILO, and you should be set. The only issue right now is resyncing swap space -- don't do it while the swap is in use. I believe there is a how-to/faq on this somewhere out there (tm). -- Randomly Generated Tagline: The argument against using an operator for other than its primary purpose strikes me the same as the old argument that you shouldn't have sex for other than procreational purposes. Sometimes side effects are more enjoyable than the originally intended effect. -- Larry Wall in <[EMAIL PROTECTED]>
Re: RAID 1 SETUP redhat 6
On Tue, Jan 25, 2000 at 04:02:25PM +0500, [EMAIL PROTECTED] wrote: > Now how will i mirror the two so that my data dont get loss. i have downloaded > the raidtools-0.90rpm. As i read the raid documentation i found that you have to > mark the parttion to fd but to my surprise my fdisk shows that there is no > partition type of fd type. fd isn't taken so it's not shown in fdisk. What I would do in your situation is to recompile the kernel w/ RAID support, and make a raidtab w/ the currently used disk as "failed-disk" (so that you can make the array but not affect the currently used disk at all). After you make all of the arrays you want, do a mke2fs on the partitions of the new RAID array. Then make a boot disk w/ the raid kernel, and then reboot using a RAID-enabled RAID rescue disk (RH6.1 works for this I believe). You can then mount the RAID array and the normal "old" partition, and copy all of the files over to the RAID array ("cd /olddir ; tar cfp - . | ( cd /newdir ; tar xpf - )"). After this is all done, you should be able to reboot with the boot disk you made w/ root set to "/dev/md0". If this works, make sure everything still works for a while. If it does, you do a raidhotadd for all of the old partitions into each of the respective RAID arrays, and let them sync up. You can then install LILO onto the device and that should be the end of it. I kind of jsut ran through a "probably would work" proceedure, but it should give you the idea. -- Randomly Generated Tagline: "There are two kinds of security: the one that will keep your sister out, and the one that will keep the Government out." - Bruce Schneier
Re: Redhat Raid1 setup/usage question
On Mon, Jan 24, 2000 at 03:26:53PM -0500, Jeff Howard wrote: > 2. When I shutdown, connect the second disk back and start up again, > the second disk doesn't seem to re-sync. I get a message that the first > disk is running in degraded mode because there's no spare disk to > reconstruct the array. I assumed that connecting the second disk > would cause the system to re-sync the disks and continue on as a > mirrored system again. > > So... does RAID1 work differently than I expected? Is there a patch > I'm missing? Do I need a third disk to actually have the system repair > itself? The RAID system shouldn't just assume that the second disk should be used to resync the RAID array. You have to do a raidhotadd to add the specific partition back to the array, which will then resync. If you had specified (or raidhotadded) a spare, then that disk would auto-reconstruct. -- Randomly Generated Tagline: "I once witnessed a long-winded, month-long flamewar over the use of mice vs. trackballs...It was very silly." (By Matt Welsh)
Re: re-construction speed
On Thu, Jan 20, 2000 at 02:52:31PM -0800, Michael wrote: > Is there a way to tune this?? so that it has more cpu time available > -- or whatever it needs?? > what is the purpose of and how do you use > > /proc/sys/dev/md/speed-limit If I remember correctly (it's been a while), the number runs from 0 to 1000, determining the reconstruct speed. Try something like "echo 500 > speed-limit". -- Randomly Generated Tagline: "Nothing takes the taste out of peanut butter quite like unrequited love." - Charlie Brown
Re: speaking of moving raid disks
On Thu, Jan 20, 2000 at 02:26:31AM -0500, James Manning wrote: > with a simple cp. I'd probably not bother with sep. ones for each array > simply b/c the "cp /proc/.../raidtab /etc/raidtab" option would go away, > although if you want to do both that'd certainly be helpful. Well, you could always do "cat /proc/.../raidtab.md* > /etc/raidtab" ... Same idea. -- Randomly Generated Tagline: "What you end up with, after running an operating system concept through these many marketing coffee filters, is something not unlike plain hot water." (By Matt Welsh)
Re: No spare disk to reconstruct array! -- continuing in degradedmode
On Tue, 4 Jan 2000, Innovation Strategies wrote: IS> How can I reconstruct my RAID1? If you're using the new RAID code (which you should be), you should be able to just "raidhotadd /dev/hda5". It will add back into the array and auto-reconstruct. -- Randomly Generated Tagline: "The only way you'll get me to talk is through slow painful torture, and I don't think you've got the grapes."- Stewie on Family Guy
Re: Large files 2GB+ & RAID?
On Tue, 28 Dec 1999, Hunter Matthews wrote: HM> Basically, nobody is going to use a true logging filesystem these days - HM> the databases themselves do rollback. The other advantage to logging is HM> the part that is also a part of journaling, and journaling is thus more HM> general purpose. [Pleading for mercy, I'm not a heavy DB guy] Slight modification for your statement: "... nobody is going to use a true logging filesystem for databases these days ..." Logging file systems are still VERY useful for general purpose FSes. -- Randomly Generated Tagline: Book never written: "Dog training." by Wille Bite
Re: Software Mirroring in Linux
On Wed, 22 Dec 1999, Craig Mckenna wrote: CM> a mirror. The problem we have is that whenever the server is restarted. The CM> Mirror comes up and the only data present is CM> the original contents of the drive, no changes to files or directories on CM> the Mirror are stored. The first thing that comes to mind is: Are you mounting the mirror (/dev/md0) or are you mounting a partition directly (/dev/sda1)? -- Randomly Generated Tagline: "MSDOS didn't get as bad as it is overnight -- it took over ten years of careful development." - [EMAIL PROTECTED]
Re: how to hotadd a spare?
On Mon, 29 Nov 1999, Markus Schulte wrote: MS> is there a way to hatadd a spare to a running raid5? After you get the box to recognize the disk, I'm fairly certain you just do a raidhotadd. It'll auto-add as a spare. -- Randomly Generated Tagline: "These periods are always 15 minutes shorter than I'd like them, and probably 15 minutes longer than you'd like them." - Prof. Van Bluemel
RE: Booting from raid1 - halfway only
On Mon, 22 Nov 1999, Dirk Lutzebaeck wrote: DL> "/boot=/dev/md?" line which was suggested in an earlier thread? I DL> don't see how this works. Going back to Dirk Lutzebaeck's problem DL> with this, he used "boot=/dev/md0". /dev/md0 is a mirror of DL> /dev/sda1 and /dev/sdb1. Why then does lilo try to install to DL> /dev/sda and /dev/sdb, not to /dev/sda1 and /dev/sdb1? because the system doesn't boot off of the first partition, it boots off of the boot block on the disk (ie: the first part of /dev/sd[ab]). ie: lilo is trying to be smart about this. (on an aside: I wonder if this would also work with a RAID-[45] array... if one disk fails, it could still boot in degraded mode. have to try it one of these days.) -- Randomly Generated Tagline: quit When the quit statement is read, the bc processor is terminated, regardless of where the quit state- ment is found. For example, "if (0 == 1) quit" will cause bc to terminate. (Seen in the manpage for "bc". Note the "if" statement's logic)
RE: Root RAID and unmounting /boot
On Wed, 27 Oct 1999, Bruno Prior wrote: BP> (a) Getting the latest patched lilo from RedHat or applying the lilo.raid1 patch BP> and rebuilding it yourself (if /dev/md0 is RAID-1) BP> (b) Providing lilo with the geometry of one of the devices in the array (again BP> if /dev/md0 is RAID-1) BP> (c) Using a slightly-adapted grub instead of lilo (again if /dev/md0 is RAID-1) BP> (d) Making sure the files to which these lines point are not on a software-RAID BP> array. Just a note: I setup root RAID1 over the weekend on my RH61 box. The configuration file is really simple, you run lilo as normal, it writes the boot block to all disks in the array and you're done: boot=/dev/md0 map=/boot/map install=/boot/boot.b timeout=50 default=linux image=/boot/vmlinuz label=linux read-only root=/dev/md0 -- Randomly Generated Tagline: "How should I know if it works? That's what beta testers are for. I only coded it." (Attributed to Linus Torvalds, somewhere in a posting)
Re: UNEXPECTED INCONSISTENCY
On Mon, 25 Oct 1999, Aaron Hatfield wrote: AH> UNEXPECTED INCONSISTENCY: RUN ; fsck MANUALLY AH> /dev/md0: The filesystem size (according to the superblock) is 1542208 blocks AH> The physical size of the device is 1542144 blocks AH> Either the superblocks or the partition table is likely to be corrupt! AH> AH> I get this for all md devices after I fsck the device manually I am able to AH> mount and use them but the next time I reboot it starts over again. looks like someone created a RAID array after creating the filesystem. oops. the only way I know of to clean this up is to create an array properly (ie: mke2fs AFTER the mkraid), then copy the data onto it. -- Randomly Generated Tagline: "A word to the wise: a credentials dicksize war is usually a bad idea on the net." (David Parsons in c.o.l.development.system, about coding in C.)
Re: 71% full raid - no space left on device
On Thu, 14 Oct 1999, Thomas Davis wrote: TD> I don't know of any Unix FS with dynamic inode allocation.. Is there TD> one? fyi: I know WAFL (NetApp) can do it, but it's not a UNIX fs... -- Randomly Generated Tagline: If it wasn't for Plumbers, you'd have no place to go !
Upgrading a RAIDed System
I currently have a server which I'm planning to setup using RAID -- not a big deal. The only thing I haven't quite figured out yet is how I will upgrade the OS once the root fs is on a RAID array ... I'm currently using RH 6.0, and am planning to either have a /boot RAID-1 setup w/ the rest of the root fs on RAID-5, or just RAID-1 the whole thing (not sure how many drives I'm going to have yet.) How does that work? I typically just pop in the CD and select "upgrade". Does anyone know if RH can upgrade a RAID installation, or will I have to come up with some other tricky way of doing it? (re-install and RAID each time, etc.) Thanks. -- Randomly Generated Tagline: "If you want to travel around the world and be invited to speak at a lot of different places, just write a Unix operating system." (By Linus Torvalds)
Re: e2fsck not correcting RAID-5 recovered filesystem
| good. I suppose that 1% is due to the filesystem data getting corrupted | due to the double-disk failure. just an idea for you guys if you're daring and have the right parts: a few months ago, I had a double-disk failure on a RAID4 array (happened on power-up). we ended up taking one of the failed disks and the known good spare disk and swapped the controllers on the drive. put the failed disk back in the array, it spun up, and we ran in degraded mode until we got a new spare to reconstruct to. you would need identical drives for this, and typically an appropriately sized torx driver, but ... depends what the drive's problem is too. -- Randomly Generated Tagline: "Money is better than poverty, if only for financial reasons." - Woody Allen
Re: partition size limit
| I think you meant 'if you're NOT really concerned about performance.' | The benchmarks I've seen for array controllers come nowhere near the | performance of a software-based array on a simple SMP server. For one | thing, the processor on, say, a DPT or ICP RAID controller is nowhere | near as powerful as an Intel Celeron, much less a DEC Alpha. Comments? w/ software RAID on a PII 233, the load on the system was hovering 2-6 (heavy writes to a RAID 5 array). w/ the hardware array, the load is <1 all the time, and the system performs (as expected) much much faster now. for that box, it's no contest for me. ymmv. btw: as long as the cpu on the raid controller is fast enough to keep up with the data you're sending, who cares if it's slower than an alpha? also, some controllers (the ones I've used anyway) have NVRAM on-board to speed up writes -- something the software raid doesn't do. -- Randomly Generated Tagline: "When you're done developing software, all you have are empty pizza boxes ..." - Prof. Michaelson
Re: partition size limit
|Has anyone tried this yet? Will things like fsck still work on a |filesystem this large? sure, but it'll take a REALLY long time. I believe that reiserfs is out now, you may want to look into that. (it's journalled as I remember, so fscks (unless forced) aren't necessary.) optionally, ext3 or XFS (both when available) would be good too. |I'm able to get the 50gig disks rather cheaply (about US$1150) so this |seems like a really inexpensive way to build a huge file store without |paying a huge Network Appliance/EMC type of premium. The system is going Yes and no. NetApp's (what I'm used to) get you more than just a big disk. I assume EMC/etc do the same. (optimized performance, journaled fs, snapshots, easily growable fs, security (they only do one thing), etc, etc.) It depends what you want to do, and how much you're willing to pay though. If you only need to give a large amount of file space to a handful of systems, local disk is the way to go. If you want to serve data to a network, you may want to plan long-term and go with something beefier and optimized for networking file serving. I have 4 NetApps which serve out data required by all hosts on my network... Wouldn't dream of trading them in for an equally sized RAID array. I also have several systems which have a need for large amounts of disk space for themselves. I wouldn't dream of trading in the RAID arrays for anything else. |to be backed up daily. I'm more concerned with total space and |read/write performance than redundancy (which is why I'm using RAID 0 |instead of 3 or 5). be warned, the MTBF of a RAID 0 array is (MTBF of single disk)/(# of disks). you're much more prone to lose all of the data that way. to back up 200Gb on one of the netapps I have, it takes 27 hours (streaming to DLT4000's). restore time is probably at least that long. I would urge RAID 5 for something that large. it's slower writes (and a smaller array) than raid 0, but when the first disk fails and you've saved yourself hours of restores, you'll be happy. and if you're really concerned about performance, I'd suggest a hardware-based RAID setup (either RAID controller or external RAID enclosure). it also makes administration a little easier (don't have to muck around with how to boot, etc, etc.) I've had great successes with the PowerRAID stuff from Zzyzx (www.zzyzx.com). |Any advice would be greatly appreciated. you haven't mentioned what you need the 300gb of space for, so advice may or may not be useful. what are you trying to do?
md or dump?
I've recently been able to procure a hw raid enclosure for a news server, and wanted to move the current spool (located on a sw raid5 array) to the new array. Seeing the horrid performance of copying between the two (~1Gb every 30 minutes), I thought it was due to the 'cp -a' I was using to copy the data. I installed dump/restore, and found that they wouldn't work at all: (/mnt1 is the sw raid array mounted ro. it didn't work in a dump | restore mode, so the examples below are trying to dump to /dev/null ...) # --- Start --- # [news ~]$ dump 0sdbf 10 10 128 /dev/null /dev/md0 DUMP: Date of this level 0 dump: Mon Mar 22 19:25:36 1999 DUMP: Date of last level 0 dump: the epoch DUMP: Dumping /dev/sda1 (/) to /dev/null DUMP: mapping (Pass I) [regular files] /dev/sda1: Ext2 inode is not a directory while mapping files in dev/md0 [news ~]$ sudo dump 0sdbf 10 10 128 /dev/null /mnt1 DUMP: Date of this level 0 dump: Mon Mar 22 19:25:40 1999 DUMP: Date of last level 0 dump: the epoch DUMP: Dumping /dev/sda1 (/) to /dev/null DUMP: mapping (Pass I) [regular files] DUMP: mapping (Pass II) [directories] DUMP: estimated 117 tape blocks on 0.00 tape(s). DUMP: dumping (Pass III) [directories] DUMP: dumping (Pass IV) [regular files] DUMP: DUMP: 201 tape blocks on 1 volumes(s) DUMP: Closing /dev/null DUMP: DUMP IS DONE [news ~]$ cd /mnt1 [news /mnt1]$ dump 0sdbf 10 10 128 /dev/null . DUMP: Date of this level 0 dump: Mon Mar 22 19:25:51 1999 DUMP: Date of last level 0 dump: the epoch DUMP: Dumping . to /dev/null .: Attempt to read block from filesystem resulted in short read while opening filesystem # --- End --- # Is this a dump problem, or a md problem? -- Randomly Generated Tagline: In plumbing, a straight flush is better than a full house!
Re: hardware RAID
| I would be quite interested to have some answers about this matter too. | (I am interested on RAID-1 too). if you want to do a form of hardware raid, you might also be interested in a RAID enclosure instead of a RAID controller. I've been using some from a company called Zzyzx (http://www.zzyzx.com/) that work very nicely. -- Randomly Generated Tagline: "You can see if I trip on the cord, or more likely WHEN I trip on the cord..." - Prof. Farr
Re: RAID monitor?
| I just wrote the enclosed. | open(IN,") { | if ( $_ =~ /\s+active\s+/ ) { | $_ =~ /\[(\d+)\/(\d+)\]/; | $count = $1; | $active = $2; | if ( "$active" ne "$count" ) { | @F = split(/\s+/); | print "Warning: /dev/$F[0] has a dead partition!\n"; | print "$_"; | } | } | } this is what I was wondering about ... From looking at a number of different lines in mdstat for different array types, etc, there doesn't necessarily need to be that "[#/#]" section. (for instance, raid0 & lvm stats don't have it) I ended up using the previously posted "diff" script (check to see that current mdstat looks like previous "good" mdstat). Simple, but it works to let me know if something changed. thanks to all btw. 8) -- Randomly Generated Tagline: "... although it's better if you call it an osculating circle because nobody knows what it means. Except those smarty-pants math professors..." - Prof. Farr
RAID monitor?
Just curious-- has anyone written a generic "monitor" script that checks the status of the RAID arrays currently in use? something that might mail if there's an error? I started working on one, but don't want to reinvent the wheel. -- Randomly Generated Tagline: If it's useless, it will have to be documented.
Re: most RAID crashproof setup = BOOT FROM FLOPPY DISK
| but do you think that there is a possibility that the disk gets corrupted at a | point which, | LILO begins loading itself (prior kernel loading) and then stops due to disk I/O | error ? usually the whole disk will just fail first. how about a compromise here: do the LILO thing (since it's not likely that it will die), and if it does, you have a bootdisk ready to go. either way, there will have to be some human intervention if the first/default disk fails. -- Randomly Generated Tagline: All warranties expire upon payment of invoice!
Re: eide raid5?
| Hmm. I use SCSI on high-performance systems, but if IDE is so bad, why | does NASA use IDE? ;) as far as I know, beowulf tends to use the network more than the disk, so it isn't necessary to have an extremely fast disk subsystem. RAM, CPU, and network speeds are much more important. -- Randomly Generated Tagline: If ignorance is bliss, why aren't more people happy?
Re: Software Raid5 configuration
|raid0 configuration currently. I would like to do away with this |configuration, and setup a linux software raid. I would prefer to use a ok |stable 2.0.X kernel but will use a 2.1.x kernel if neccessary. I am shouldn't be a problem unless you need some of the 2.1.x (or now 2.2.0preX) features. |curious as to whether I can set them up in the following configuration. 3 |Drives for Data 1 for parity and another for a hot spare. If so, I've read That's not a problem, but you're describing a RAID 4 setup (RAID 5 doesn't have a dedicated parity drive.) |through several raid-howto's and none of them give a step by step setup of |raid5 on Linux. Can someone send me a list of things to do to go about |setting this up, or if not does anyone know of one that's currently out |there that they could point me to? Thank you, let's see if I can remember: 1) grab kernel, apply latest raid patch + raidtools source 2) compile+install kernel, compile + install raidtools 3) reboot (load new kernel) 4) use fdisk to create partitions you want (possibly 1 partition taking up entire drive), make sure to set partition type to 0xfd (will allow auto-start at bootup). 5) create /etc/raidtab. mine looks like this, you'll have to modify it slightly if you want spare-disks (as above): raiddev /dev/md0 raid-level 5 nr-raid-disks 4 nr-spare-disks 0 chunk-size 256 parity-algorithmleft-symmetric persistent-superblock 1 device /dev/sdb1 raid-disk 0 device /dev/sdc1 raid-disk 1 device /dev/sdd1 raid-disk 2 device /dev/sde1 raid-disk 3 I believe you just up the nr-spare-disks to 1, and add in a spare-disk command at the end, but look in the man page, it'll tell you. 6) (at this point my memory is a little foggy) do a /sbin/mkraid /dev/md0 this should create the md device for you. 7) mke2fs /dev/md0 8) you should be done. I'd try rebooting to make sure the auto-start works, and that's about it. you can then mount up /dev/md0 as you like.
Re: Can't mkraid
| I've created the appropriate partitions, set the partition ID to fd, and | created the appropriate /etc/raidtab file. Running: | invalid chunk size (0Kb) what does this "appropriate" raidtab file look like? you should have a line like this in it: chunk-size 256 -- Randomly Generated Tagline: 8 out of 5 doctors feel it's OK to be skitzo!
Re: help how to apply patch ?
| I am running RedHat version 5.1 (kernel 2.0.35) | and have pulled down the patch file | raid145-0.36.3-2.0.30.gz make sure you have the 'kernel-source' RPM installed (or just go grab the source from ftp.kernel.org or a mirror). if you grab the source, untar it in a directory (this is typically /usr/src). Then goto that directory, and you should see a directory called "linux". from here, do a "patch -p0 < raid0145#", where is the version and all. Make sure 1) you have an updated version of the patch, and 2) make sure the file is ungzipped when you redirect it to patch. patching isn't very hard. 8) -- Randomly Generated Tagline: "My psychic guided us to our best fishing trip ever..." - Commercial
kernel autostart not working
I reported this last week, but thought I'd send in an update. I've been running a RAID 5 array w/ kernel 2.1.125 & raid 19981005. When I upgraded to 19981105, autostart stopped working. It has continued to not work w/ versions 19981106 and 1108. To fix some fs corruption that occured last week, and to try to eliminate possible problems, I changed kernels to 2.0.35, and applied raid 19981108. I then remade the array. Here's the raidtab: # Sample raid-5 configuration raiddev /dev/md0 raid-level 5 nr-raid-disks 4 nr-spare-disks 0 chunk-size 256 parity-algorithmleft-symmetric persistent-superblock 1 device /dev/sdb1 raid-disk 0 device /dev/sdc1 raid-disk 1 device /dev/sdd1 raid-disk 2 device /dev/sde1 raid-disk 3 here are the relevant bootup messages from syslog: Nov 9 18:08:35 news syslogd 1.3-3: restart. Nov 9 18:08:35 news kernel: klogd 1.3-3, log source = /proc/kmsg started. Nov 9 18:08:36 news kernel: Loaded 4215 symbols from /boot/System.map. Nov 9 18:08:36 news kernel: Symbols match kernel version 2.0.35. Nov 9 18:08:36 news kernel: No module symbols loaded - kernel modules not enabled. Nov 9 18:08:36 news kernel: ),R:3,S:6> Nov 9 18:08:36 news kernel: D 4: DISK Nov 9 18:08:36 news kernel: D 5: DISK Nov 9 18:08:36 news kernel: D 6: DISK Nov 9 18:08:36 news kernel: D 7: DISK Nov 9 18:08:36 news kernel: D 8: DISK Nov 9 18:08:36 news kernel: D 9: DISK Nov 9 18:08:36 news kernel: D 10: DISK Nov 9 18:08:36 news kernel: D 11: DISK Nov 9 18:08:36 news kernel: THIS: DISK Nov 9 18:08:36 news kernel: md0: no array superblock. Nov 9 18:08:36 news kernel: rdev sdb1: O:sdb1, SZ: F:0 DN:0 rdev superblock: Nov 9 18:08:36 news kernel: SB: (V:0.90.0) ID: CT:364742ef Nov 9 18:08:36 news kernel: L5 S08890112 ND:4 RD:4 md0 LO:2 CS:262144 Nov 9 18:08:36 news kernel: UT:364767a3 ST:1 AD:4 WD:4 FD:0 SD:0 CSUM:b1d8da1a Nov 9 18:08:36 news kernel: D 0: DISK Nov 9 18:08:36 news kernel: D 1: DISK Nov 9 18:08:36 news kernel: D 2: DISK Nov 9 18:08:36 news kernel: D 3: DISK Nov 9 18:08:36 news kernel: D 4: DISK Nov 9 18:08:36 news kernel: D 5: DISK Nov 9 18:08:36 news kernel: D 6: DISK Nov 9 18:08:36 news kernel: D 7: DISK Nov 9 18:08:36 news kernel: D 8: DISK Nov 9 18:08:36 news kernel: D 9: DISK Nov 9 18:08:36 news kernel: D 10: DISK Nov 9 18:08:36 news kernel: D 11: DISK Nov 9 18:08:36 news kernel: THIS: DISK Nov 9 18:08:36 news kernel:** Nov 9 18:08:36 news kernel: Nov 9 18:08:36 news kernel: md0 stopped. Nov 9 18:08:36 news kernel: autorunning md0 Nov 9 18:08:36 news kernel: running: Nov 9 18:08:36 news kernel: now! Nov 9 18:08:36 news kernel: md: former device sdb1 is unavailable, removing from array! Nov 9 18:08:36 news kernel: md: former device sdc1 is unavailable, removing from array! Nov 9 18:08:36 news kernel: md: former device sde1 is unavailable, removing from array! Nov 9 18:08:36 news kernel: raid5: device sdd1 operational as raid disk 2 Nov 9 18:08:36 news kernel: raid5: not enough operational devices for md0 (3/4 failed) Nov 9 18:08:36 news kernel: RAID5 conf printout: Nov 9 18:08:36 news kernel: --- rd:4 wd:1 fd:3 Nov 9 18:08:36 news kernel: disk 0, s:0, o:0, n:0 rd:0 us:1 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 1, s:0, o:0, n:1 rd:1 us:1 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 2, s:0, o:1, n:2 rd:2 us:1 dev:sdd1 Nov 9 18:08:36 news kernel: disk 3, s:0, o:0, n:3 rd:3 us:1 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 4, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 5, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 6, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 7, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 8, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 9, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 10, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: disk 11, s:0, o:0, n:0 rd:0 us:0 dev:[dev 00:00] Nov 9 18:08:36 news kernel: raid5: failed to run raid set md0 Nov 9 18:08:36 news kernel: pers->run() failed ... Nov 9 18:08:36 news kernel: do_md_run() returned -22 Nov 9 18:08:36 news kernel: unbind Nov 9 18:08:36 news kernel: export_rdev(sdd1) Nov 9 18:08:36 news kernel: md: bug in file md.c, line 147 Nov 9 18:08:36 news kernel: Nov 9 18:08:36 news kernel:** Nov 9 18:08:36 news kernel:* * Nov 9 18:08:36 news k
2.1.125/raid-19981005 ...
Just curious, There was a large rejection when applying raid0145-19981005 to the 2.1.125 kernel source (problems with drivers/block/md.c). Is there a new version being released to cleanly apply to 2.1.125, or can I just copy over the patched 2.1.124 md.c and go from there? The only difference between the kernel md.c is: *** *** 696,702 RO_IOCTLS(inode->i_rdev,arg); default: - printk ("Unknown md_ioctl %d\n", cmd); return -EINVAL; } so I would imagine copying would work fine. -- Randomly Generated Tagline: A hen is an egg's way of making another egg.