Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×
Hardware

How Are RAID Arrays Identified By Hardware? 72

Coward Anonymously Before Me asks: "This is more of a tech/hack question, but recently my highpoint controller forgot my disks were in a raid array. All the Disks still function, and have ZERO problems, aside from being not identified as still in RAID-0. All the data should still be there, but remains unaccessible to me, thus the question how and where would this kind of information be stored? On chip? MBR? and can the data be recovered without 3rd party interaction via free/open source toolkits? or even purchased software?"
This discussion has been archived. No new comments can be posted.

How Are RAID Arrays Identified By Hardware?

Comments Filter:
  • by i_am_nitrogen ( 524475 ) on Friday December 13, 2002 @11:16AM (#4880534) Homepage Journal
    ...you ought to be able to just dd the data from the drive, and if the chip doesn't use a non-standard data layout, write a program or script to put it back together and back it up, then you recreate the array, and put your data back on it.
  • Uhh (Score:1, Insightful)

    by fredrikj ( 629833 )
    Did you try Google [google.com] before submitting that?.
    • When questions like this come up in our daily tech life we don't just think "Oh! I'll goto google and get an answer!" Some of us are community conscience and think "This is an interest problem, I bet the /. community would be interested in the answer." Of course when we have these thoughts we forget how unforgiving some readers are *cough*
      • Re:vs RTFM (Score:4, Insightful)

        by Lord Bitman ( 95493 ) on Friday December 13, 2002 @12:52PM (#4881370)
        people get annoyed with me that I ask questions on IRC or message boards which are covered thoroughly in manuals. They respond with, of course, "RTFM"
        Am I the only one who thinks that a Manual is a pretty lame source of information to reach for first-thing?
        I have a few sources I go through, usually the manual is one of them, but I _Always_ first ask a person who might have the answer on hand. Manuals are not often things which lend themselves to answering typical questions such as "Can I blah?". The problem with "Can I blah" being looked up in a manual, among other things, is that often there are numerous synonyms for 'blah', and only one of them is ever used in the book, especially the index.
        Perhaps they mean "Read the entire manual before even using the product". The obvious problem here is that manuals are getting longer every day. I've heard that some Linux Distro comes with a 2000 page manual just for getting it installed. Obviously, to read an entire manual before using a product would leave little time for using any products, and leave you more or less unknowledgeable about the product.
        Then there's the problem of phrasing. Manuals may answer your question, but only burried in a lot of other information which isnt related to what you're actually working on. A person who knows already, however, can simply answer your question.
        Slashdot, however, is far too public and non-specific. There's no reason to ask this kind of question on slashdot, get some friends or something.
        • That's the sort of attitude which makes people prefer to run Windows. I can find the answer to *any* windows problem I have within ten minutes on google - Linux isn't there yet. So surely you should be helping people?
          • I accidently deleted my first reply to this so I'll make this one brief. It seems there are just as many, if not more, hits on google that deal with Linux than Windows (about 54 vs 59 million for linux vs microsoft. accounting for windows pages not dealing with computers, I'd say they're about even). There are certainly more linux "answers" out there than windows, its just that most of the linux answers that you need are probably basic enough that most of the people running linux don't ask them. Windows is incredibly difficult to find information for. Once you get into hardcore networking and stuff like DCOM and intricate implimentations thereof, you're out of luck in the windows world.
          • That's strange.

            It's been my experience, especially with simply coding questions, that it's easier to deja and google for competent Linux/Unix answers than it is for competent windows answers.

            Of course, I often find more windows answers overall, it's simply that they just aren't very helpful [driverforum.com].
          • That's also the sort of attitude that makes sane people dread writing software for Windows. Especially for the really odd problems, half the time you'll find other people on Google asking the same questions, and no signs of any answers.
          • I have a *far* easier time trying to find answers for linux. If a google search doesn't do it for me, there's almost always a mailing list I can join and talk ***directly to the developers*** There is nothing like that in the Windows world. I've also had much better luck finding code/programming examples for Linux. I recently was trying to work with Audio APIs in both OS's and had much better luck with Linux until I found some open source cross-platform code to use w/windows.
        • The problem with that is that there are very few people who'd like to answer beginner questions. Most people who come to those channels to help have a good knowledge and want to feel useful. I don't know about you, but I wouldn't come to a mathematician to ask her how to compute the square root of something. Surely she's got something much better to do.

          This is why things like FAQs exist, people are tired of answering the same questions over and over, when there are far more interesting things to do. #debian has the apt bot, for example.

          Also, it depends on how you ask the question. If you think that people in support channel come there for answering every of your questions, you're mistaken and will be rightfully flamed.

          BTW, manuals have an useful feature called "index". Maybe you could try using it.
          • BTW, manuals have an useful feature called "index". Maybe you could try using it.

            Grandparent wrote that many manuals aren't well-indexed: "The problem with 'Can I blah' being looked up in a manual, among other things, is that often there are numerous synonyms for 'blah', and only one of them is ever used in the book, especially the index" (my emphasis).

            • Of course not every manual makes it easy, but there are some that are made exactly for that. For example, the Perl Cookbook. The whole book is just "How do I" questions and answers.

              About the synonyms problem. Some books I've got here have a short introduction in the form of "foo is also known as bar and baz, but for consistency reasons we'll call it foo". If it's hard to find information in a book then it's just not worth buying. And anyway there's always Google.
              • And anyway there's always Google.

                But not everybody knows the secrets of forming a good Google query. Google is good, but not good enough to read your mind... yet.

                • True, but if you can't find it in Google then I doubt anybody will tell you to RTFM. If somebody asked me how to burn CDs in Linux, I'd tell him/her to RTFM because there's a lot of documentation available, and google returns plenty of it on the first results page.
                  • If somebody asked me how to burn CDs in Linux, I'd tell him/her to RTFM

                    On the other hand, I'd help the user learn to formulate a query: "Go to Google.com, enter [ linux burn cd ], and click Google Search." Or, if a question is answered in the FAQ: "Look in the DJGPP FAQ, section 8.3."

                    • I explain to people how to burn cd's every day, some of them can't figure out something that technical on their own. Other's know that I can show them in 5min what would take them hours to figure out on their own. All of them know I'm busy and understand I'll get there whenever I can. Yes it's frustrating to answer simple questions, but there is a big difference between "how do I click a mouse?" and "how do I burn a cd?" The person who is asking how they burn a cd, has at least done enough looking around to find out that what they want to do is burn a cd.. maybe they are trying to reach a different end goal and burning a cd is one of the things they want to try along the way. Maybe their really trying to transfer files from one machine to another or find a reasonable way to share bulk files with friends and have just spent a few hours trying to research how to get to the point where they've got those bulk files (perhaps a family history album, this is recent real world example I helped someone at this point with). Just because the question they ask has a simple solution for someone who has already figured it out, or has another possible source of information, doesn't mean they haven't tried to figure something out for themselves.

                      Maybe they don't know there are such things as search engines yet, maybe what they really need help with finding that information out, perhaps along with the answer to their original question you could explain search engines, I'm sure they would really appreciate that.

                      People often have different reasons, they may just have a different knowledge set than you. They may not be as intelligent, and where you were able to interpret the information you found during a google search easily, they need someone to translate it into something a little more user friendly before they will understand it. Perhaps as I said earlier, they are frustrated and need a simple answer before they can move back to tackling their real issue.

                      I don't know about you, but I work in a tech shop, all day long we ask each other questions, a great deal of them simple, it's only when someone else isn't available or the others don't know that you resort to a web search that takes more time. I'll evaluate how important it was for you to spend 5 minutes to answer a question I would have had to spend 2hrs figuring out on my own when you need to know something that I have experience with and could answer in 5mins and will save you 12...

                      Volunteers can do whatever they wish, be as friendly or not friendly as they wish, they can cater to whatever technical degree they wish on a given subject... but they certainly have no reason to flame those who have simpler questions, if you tired of it, or would rather help with more difficult things, so be it but that's an excuse for not answering it's no excuse for "RTFM".

                      I will say this though, I've spent a great deal of time on help channels.. both giving and recieving information. One, if you give information where you can, your more likely to get the information you need without a hassle. And two, if you tried to find the information and couldn't, someone will usually point you to it, even if they do so in the form of a flame. And three, if you did find the information, are bright enough to mention what you weren't clear on, someone will usually phrase it in another way or otherwise clarify even if it is in a exclamation that ends in "you fucking n00b".
                    • Anyone know where I can find information on the Rockridge standard? I can't find it anywhere and I heard it can burn files to CD including permissions information, I need to get specification though since I need to have each .iso generated by a script..
                      What? I figure we're off-topic enough already as it is :D

                      Though it does bring up an interesting point that some people really just need to know where a manual is, or if one exists. Manpages often tell everything you could ask for regaurding command line switches, but doesnt talk at all about the interface, or even what it is the program does. The most helpful source of information could be burried in a seemingly unrelated HOWTO which you'd never know to search for without knowing the program's intended function in the first place [you just found it in some error report from cron]
                      I get the feeling that a lot of people who say "RTFM" havent actually read the specific manual in question, they've learned from others, and through experience with related things. I hate being told to RTFM after I just went through the entire thing and have found nothing even related to my problem.
        • given that a large portion of the documentation for a distribution is available on the computer, one might try searching through the documentation with the computer, if the index is not sufficent.

          man and info both support searching (with the / key in both, i believe)

          most of the documentation for linux/unix that is not in man or info pages is in html, and most web browsers also support searching, although the key may vary

      • Re:Uhh (Score:2, Insightful)

        by bootprom ( 585277 )
        If everyone was like you, people who really needed help with issues that were *not* in the FM would never get it. Forums like usenet, irc, and even ask /. are there to discuss novel issues, not to re-hash old topics over and over. People who ask questions before they RTFM are a detriment to the online community. We are *all* fortunate that people put with the likes of you.

        When you *do* RTFM, not only do you get better at manual reading, but you also pick up additional information, so you may be less clueless the next time you have a problem

        -BP
        • your issue really shouldn't be with me (cliff chose to post the question).

          I'm trying to give a little perspective to the issue once a question has been accepted. Flaming the person who asked doens't do any good. /.'s staff made the choice (supporting my perspective) that the /. community would be interested in the discussion of this question. *Note* that it didn't make the front page of /. Infact it's in the "ask /. section which is designed for this very type of discussion. So if you want to Flame Flame /. for not designing thier website to YOUR interests

          • Your right, I had *meant* to flame (bit harsh of a word I think) Lord Bitman, but I got sloppy with the mouse. My apologies.
            • wow that sheds a whole new light on your post LOL...sorry if I sounded bitchy...I've now had a good laugh for the day thanks.....Whee hee he he heeee. Company christmas party tonight Free (as in Free BEER) BEER.
      • Communities tend to take too long to get back to you because we are all buys working. Believe it or not, you are probably not the first person to have the problem you are looking for. By going to Google Groups you are taking advantage of the fact that those people who had a specific problem (most often *not* in the FM) have already discussed and solved the issue. This is the last place I would look for very specific configuration solutions.
  • Space (Score:4, Informative)

    by Hank Reardon ( 534417 ) on Friday December 13, 2002 @11:20AM (#4880579) Homepage Journal
    I'm most familiar with SCSI RAID, so YMMV with IDE variants. I'm not too sure about exactly where the information is, I've never been able to get rid of the RAID marking without using the drive utilities.

    Typically, there's a utility in the RAID configuration that stamps the drive as part of a set, marks the state (good, bad, rebuilding or hot-spare are most common) and some kind of versionig information.

    I've ripped a few disks out of the array, mounted them as standard, reformatted and replaced the MBR, threw them back in the array, and still had them recognized as part of the RAID volume. The RAID card didn't like this much, however. :)

    I think your best bet is to talk to one of the people who actually wrote the drivers for the card (you've got the Linux source, right :) or possibly see if you can get ahold of an engineer at the manufacturer and discuss ways of getting the information back.

    Good luck!

  • Google.... (Score:2, Informative)

    by MImeKillEr ( 445828 )
    ...I went to Google and found this:

    "Where does the controller store its configuration? Disk, Ram, Rom ? Can it be saved or exported as a backup?

    The config is in NVRAM on the controller, but there's also about a 4MB
    "partition" (not really a partition, but config area) on each drive in the
    array that stores information on things like position in the array, etc.

    Alternately, try this link [makeashorterlink.com] for the whole thread on the subject.

    It literally took me 30 seconds to find it. Next time, try Google before posting, mmkay?

    • I went to Google and found this ... It literally took me 30 seconds to find it.

      Next time, if you're reporting a Google result, please give the query string that you used. This way, people who read your comment become better users of the Google search engine.

      • That's easy -- Simply CLICK on the link. Sure, its routed through makeashorterlink, but its not like I'm redirecting to a fscking goatse site.

        I put it into makeashorterlink by habit.
  • Depends (Score:4, Informative)

    by duffbeer703 ( 177751 ) on Friday December 13, 2002 @11:50AM (#4880857)
    Logical volume managers (AIX and Veritas anyway) store a unique ID on the disk, and then keeps track of what volumes are there, how they are configured, etc.

    Hardware controllers generally reserve a small slice of disk to store configuration data. Sometimes this slice is marked unusable and can only be accessed by low-level hardware.

    One of the big, unadvertised problems with RAID, particularly with new/buggy controllers, is that a controller failure can trash your data.

    Unless you have the time & knowledge to reconstruct the data structures, a controller failure that screws up the configuration data on disk effectively destroys your data.
  • on me, but only after about 3 years of continuous service (pretty good for inexpensive consumer hardware anymore). The only solution worth my time? Another HighPoint controller. Anything else will make you cry and wished you just shelled out a few bucks in the first place.


    Unless I misunderstand - your system *is* broken without the controller anyways right? So why not just replace it?

    • I went through this whole thread and -- being a former IT director that has dealt with 3 dead RAID controllers over the years -- you are by far best off with the above solution.

      In other words: Get another HighPoint controller, reconfigure it the same way the old one was, and see if it works. If it doesn't, you are most likely screwed and will need to restore from backup. Good luck.
  • Call tech support. (Score:1, Interesting)

    by draziw ( 7737 )
    Slashdot is not Highpoint Technical Support. Your question isn't how a software only IDE RAID stores data, your question is how do you fix your Highpoint config.

    You can reach the web page at highpoint-tech.com [highpoint-tech.com], and reach support via support@highpoint-tech.com [mailto]

    Good luck to you, but ask your question in a support fourum.
  • Ask Highpoint (Score:4, Informative)

    by MrResistor ( 120588 ) <peterahoff.gmail@com> on Friday December 13, 2002 @01:51PM (#4881911) Homepage
    If your controller just spontaneously lost it's configuration, it's a problem with the card. Call the manufacturer and get a replacement. If they will configure it for you before they send it (which shouldn't be difficult for them), you should be able to just swap it in and go.

    You'll need to do something like this, since you won't be able to get at the data on the drives unless you can hook them up to a RAID controller that will recognize the particular flare code etc of your setup.

    If you can stomach losing the data (you backed up the important stuff, right?) then you could try starting over from scratch, but I would not trust your RAID controller if I were you. Replace it or don't use it.

  • A bit of info (Score:4, Informative)

    by jemhddar ( 53448 ) <matt,helling&gmail,com> on Friday December 13, 2002 @01:52PM (#4881918)
    I've worked quite a bit with AMI controllers and Adaptec. At work we looked at using an Adaptec ZCR card but chose not to for the following reason.

    AMI Megaraid(and now LSI) write a bit of config info to each disk and to the controller. On these cards, you need to know the drive designations (which is drive1, which is drive 2, etc and the stripe size (how much data to write to the first disk before moving on to the second). On these controllers, if your card goes belly up you can usually put in a new card which will detect that your drives still have a configuration and use it. Otherwise, you can create a new configuration of the drives (same raid level, same stripe size, each drive with same designation) and it will access the data just fine on a reboot. (probably 75-85% of the time. The rest of the time you are just SOL and need to get out tape.)

    Adaptec's ZCR card we were testing and going to ship had the unfortunate effect that when an array was created, it immediately initialized (format) all the data.

    This is something you should check into. Perhaps the highpoint card will let you make a new array and reboot. Or it might automatically initialize and wipe out all your data before letting you use it.

    Hope this helps

  • I had a very similar thing happen to me a while back. It was caused by a hardware problem (I was using a pair of IBM deskstar drives), so I never did recover quite all the data, but I did manage to get the RAID array back together. By a lucky coincidence, this happened shortly after drivers/ide/hptraid.h was added to the linux kernel, so I had somewhere to look for inspiration.

    This file describes the structure of one sector somewhere near the start of each disk. (Sorry, I don't remember exactly which one.) The magic number had changed on one of the disks from HPT_MAGIC_OK to HPT_MAGIC_BAD. Editing it back again was sufficient to reconnect the drives.

  • When moving an array to another system that doesn't recognize it, I usually just "add array" and enter the parameters of the config. Then reboot and all of the data is intact.
  • Is this a HPT370 based controller? I had a very similar thing happen to me a year or two ago. The RAID controller just "forgot" its stripe config, which I think is stored on a very small reserved segment of the disc. Anyway, there's a utility called raidrb (google on raidrb.zip) that I used which fixed the problem without data loss. This used to be covered in an excellent FAQ that was on viahardware.com, but apparently it's gone now.
  • Low level the drives using the RAID controller card, recreate the array, and restore from last night's backup.

    Voila! Complete Recovery in less time than it took for all these responses to be posted.

    No backup? Ewww - you are going to hate Monday.

  • 1) Backup the drives manually just in case (copy the raw drive data off to tape after booting from CD or floppy).

    2) Go into the hardware raid setup and set the array up exactl as before, but dont let it initialize. If your hardware raid controller always initializes new divces by writing over them with zeros or something, this might be undoable or tricky. If you end up initializing, it's not that big a deal.

    3) If you did it without initializing, you're probably good to go. If you ahd to initialize in order to configure the array, now boto off of CD again and restore the raw drive images from tape.

    Done
  • Answer: Most RAID (SCSI and IDE) place configuration information at the end of the drive in the array. Typically one or more logical cylinders are lopped off the drive size and used to store the config info. As the config info is not part of the drive's "size" when in RAID mode, the info is inaccessible. Also, the info is drive specific, so copying(dd, ghost, etc) won't solve the problem completely. Don't expect anything out of the config info as it is VERY manufacturer specific.

    Solution:
    1. Backup everything.
    2. Copy (dd, ghost) the most current drive to the least current drive.
    3. Go into the Highpoint BIOS config and re-declare the drives as a RAID-0.
    4. At this point the Highpoint BIOS will probably try to initialize the RAID-0. Since you copied the drives using a sector copy program in step 2, the direction of the Highpoint initialization doesn't matter.
    5. If this worked, weverything will boot normally with the RAID-0 config.
    6. If this didn't work, the drives are probably irrepairably hosed. Restore from the backup you made in Step 1. You did make a backup in Step 1, right? :-)

    Good Luck,
    Ed

  • About a year ago the exact same thing happened to me with the built-in HighPoint HPT-370A controller on my motherboard. I was able to recover all my data by simply recreating the array exactly as before, then running an "FDISK /MBR" (I'm uncertain what the Linux equivalent is). This 'undeleted' the primary partition. I then used a program called "testdisk" (try Google), which runs under DOS, Linux, and several other OSes, to undelete the other partitions.

    YMMV, but good luck.
  • Did you actually say that you're willing to purchase software in order to perform a task that is worth quite a bit to you?

    Someone alert the FSF! You're clearly insane! The software industry's brainwashed you, and you need help!
  • The real question you should be asking yourself is, of course, "Where are my backups?"
  • I'm not familar with the consumer IDE raid stuff although I've worked quite a bit with Compaq SmartArray controllers.

    Infact, I have a Compaq Smart array SCSI RAID controller in both my PC's at home (desktop and server). They're available on ebay for very nice prices and are well supported by Linux.

    The reason I mention it, is that I can pull all the disks out from one machine and pop them into the other - I don't even have to keep them in the same order! - and the controller is able to read the array configuration data from the disks and the logical volume(s) are immediately accessible. It's pretty sweet.

    If it's just your RAID controller that's failed, try putting the disks into another machine with the same type of controller, or try replacing the controller. I'm not making any promises, but if it's a decent controller it should read the saved config data from the disks and make your logical volume available.

    Have you tried the manufacturers tech support line? I'm sure they've delt with this situation before. I'm not trying to be a wiseass, but sometimes contacting the manufacturer is going to get you in touch with the folks who are most knowlegable about the product.
  • Well, it seems that a few people had smart ass comments such as "Read the Manual" or "Slashdot is not Highpoint Tech support." Well, maybe you should take your time and use it more wisely. For all you know, your "advice" has allready been attempted. If the person who posted reads slashdot, he/she more then likely knows the nature of the slashdot comunity. If thats the case, then most of your advice, such as read the manual, and call highpoint has been done. When was the last time you e-mailed tech support and got an answer back? I never have besides "We are working on a resolution to your problem..." I have too much time on my hands because I am writing telling you what assholes you are, but maybe it would be nice to help rather then refer. Also, what if the controller is onboard? For a bunch of Linux geeks, I'd expect someone to think of that.

A morsel of genuine history is a thing so rare as to be always valuable. -- Thomas Jefferson

Working...