locked
Disk 0 is 87% full & Disk 1 is 0% full = how to force duplication on Disk 1 RRS feed

  • Question

  • Hi All -

    Disk 0 is 87% full & Disk 1 is 0% full.

    Disk 1 was added today due to the recent failure of Disk 1.

    I am showing 860GB of system usage and I now believe that is folder duplication not being migrated via Drive Extender.  In my effort to resolve this I also performed a WHS reinstall.

    How do I force the move of duplicate folders to the empty drive?

    I read the Drive Extender Not Migrating Data post BUT it still unclear how to fix this.

    Please help, I have 8+ hours invested in this disk replacement.  So far no data lost, just TIME!

    Thanks - Brad

    Thursday, July 8, 2010 7:48 PM

All replies

  • 800+GB of "system" space suggests a disk problem. Have you run chkdsk on all the drives in your server?
    I'm not on the WHS team, I just post a lot. :)
    Friday, July 9, 2010 12:30 AM
    Moderator
  • Hi Ken -

    Glad you took interest in my post.

    I have not forced a chkdsk on reboot since I replaced disk 1.

    Do you recommend the chkdsk script I used when disk 1 went bad 2 weeks ago?

    You know, this one?

    Thanks Brad

    You can use the following batch file to force chkdsk on all your drives. Copy the code below and paste it into notepad, then save it (on your server) as e.g. "checkall.cmd":

    net stop pdl
    
    net stop whsbackup
    
    chkdsk D: /x /r 
    
    chkdsk C: /x /r
    
    for /d %%1 in (C:\fs\*) do start chkdsk /x /r %%1
    Friday, July 9, 2010 12:38 AM
  • Yes, that's the script I would recommend.
    I'm not on the WHS team, I just post a lot. :)
    Friday, July 9, 2010 1:03 AM
    Moderator
  • Ken -

    Before I do the chkdsk (overnight) please look at the following screen shot.

    http://www.verohomies.com/file_usage.JPG

    I am wondering if the 860GB is from files outside of DE.

    Maybe I can delete some WITH YOUR DIRECTION prior to chkdsk.

    I have yet to loose DATA!

    Thanks, Brad

    Friday, July 9, 2010 1:22 AM
  • Hi - I have a similar perplexing issue:  I have 4 drives, one is at 98% and the other 3 are about half-full - I have nearly 2 TB of storage space available!  Why isn't demigrator (which I confirmed is running) moving some data over to these free drives?  I'm concerned about the integrity of this one drive that is nearly full.

    I considered using the Drive Balancer add-in but read in these forums (by you, in fact, Ken) that use of this app is not recommended.

    Note I also got the nasty error message "The group or resource is not in the correct state to perform the requested operation" which luckily was for some shareware I was trying out so I deleted it, but this suggests either a drive failure issue or potentially something to do with this overfull drive - which was one of the drives this shareware was duplicated on - caused the problem.  So I'm running the same chkdsk batch file above to see what's what.  I ran this about a month ago and found no bad sectors.

    So I'm not sure if I have one or two issues but still want to figure out how to kick-start demigrator to distribute some of that data off my poor overfull drive.

    As a side-note, with file duplication on, how can something like this happen?  Shouldn't WHS just repoint the tombstone to the duplicated data and re-duplicate the data elsewhere?  --and hopefully report on the problem and that it resolved it?  Wouldn't that be nice?

     

     

    Friday, July 9, 2010 5:33 PM
  • How did you copy the files to the shared folders - via network/UNC path or locally to D:\shares? The later may cause duplication not to function, since in that folder are only tombstones expected, not the real files.

    Best greetings from Germany
    Olaf

    Friday, July 9, 2010 9:20 PM
    Moderator
  • Hi - I have a similar perplexing issue:  I have 4 drives, one is at 98% and the other 3 are about half-full - I have nearly 2 TB of storage space available!  Why isn't demigrator (which I confirmed is running) moving some data over to these free drives?

     

    Because it's not designed to do that.  It puts data on the drive with the least free space (leaving 20 GB of free space per drive for file expansion).

    I'm concerned about the integrity of this one drive that is nearly full.

    I considered using the Drive Balancer add-in but read in these forums (by you, in fact, Ken) that use of this app is not recommended.

    Note I also got the nasty error message "The group or resource is not in the correct state to perform the requested operation" which luckily was for some shareware I was trying out so I deleted it, but this suggests either a drive failure issue or potentially something to do with this overfull drive - which was one of the drives this shareware was duplicated on - caused the problem.  So I'm running the same chkdsk batch file above to see what's what.  I ran this about a month ago and found no bad sectors.

    So I'm not sure if I have one or two issues but still want to figure out how to kick-start demigrator to distribute some of that data off my poor overfull drive.

    As a side-note, with file duplication on, how can something like this happen? 

    You need to remove the failed drive first.

    Shouldn't WHS just repoint the tombstone to the duplicated data and re-duplicate the data elsewhere?  --and hopefully report on the problem and that it resolved it?  Wouldn't that be nice?

     

     


    Friday, July 9, 2010 10:16 PM
    Moderator
  • Hi Olaf - no I used the UNC paths.

    Kariya21 - thanks for your answer on the 98% question.  But I'm wondering about your answer to how to save data when file duplication is on and one drive is going bad (which it looks like I do have):  Once you get this error message, the source data is gone, per the forum entry http://social.microsoft.com/Forums/en-US/whsfaq/thread/7ef4db6c-4a73-42e7-b79e-cacd41365442: "This message indicates that a tombstone file is pointing to a shadow file that no longer exists on the home server."

    How, then, am I supposed to know there's a problem with one of the two drives this data is duplicated on, in order to remove the failed drive first??

     

    Saturday, July 10, 2010 12:45 AM
  • Hi Olaf - no I used the UNC paths.

    Kariya21 - thanks for your answer on the 98% question.  But I'm wondering about your answer to how to save data when file duplication is on and one drive is going bad (which it looks like I do have):  Once you get this error message, the source data is gone, per the forum entry http://social.microsoft.com/Forums/en-US/whsfaq/thread/7ef4db6c-4a73-42e7-b79e-cacd41365442: "This message indicates that a tombstone file is pointing to a shadow file that no longer exists on the home server."

    How, then, am I supposed to know there's a problem with one of the two drives this data is duplicated on, in order to remove the failed drive first??

    Unfortunately, identifying failed drives is usually much easier than identifying failing drives.  Failed drives are obvious (no response at all, WHS Console says a drive is "missing", etc.)  Failing drives can be more conspicuous (slow response times, general "odd behavior" of the OS, etc.)  The best thing to do when experiencing a hardware problem that could be related to a failing hard drive is to run chkdsk /r on each drive in the server (the FAQ post that Ken linked to above).
    Saturday, July 10, 2010 4:11 AM
    Moderator
  • So you're basically indicating that if a hard drive sector fails, the corrupted data on that sector is then copied to the duplicate shadow, making both the file and its duplicate useless.  You would think WHS would be smart enough to do a CRC check before duplicating the data!  It sounds like file duplication is only helpful in the event of a complete HDD failure, like a motor burnout - where the drive fails before the demigrator has time to update the duplicate file with the corrupted file.  Hmmph - sectors going bad is a much more common scenario, as old drives begins to deteriorate.  I feel the value of file duplication is suspect, and only not unreasonable b/c SATA HDD's are so cheap these days.

    So a nice, logical feature, if feasible, would be for the file duplicator to do that CRC check if it ever identifies a change in the file, recognize its now a corrupted file, and create a DIFFERENT original shadow file somewhere ELSE, using the uncorrupted DUPLICATE shadow.  Would that be something I could throw out as an enhancement request?

    So back to my situation:  Chkdsk found 350MB of bad sectors of a 1.5TB drive - on the D: partition of my system drive.  HOWEVER, as it is a Seagate drive I was prompted to run the "Seatools" HDD diagnostic utility before RMA, which even after its "LDST" (Long Drive Self Test) found no problems whatsoever with the drive!  This makes me believe the drive itself is actually OK.  My reasoning (and sorry to hijack this thread, Verohomie) is this:  I had virtually the exact same size of bad sectors previously on 2 separate HDD's.  Ken Warren asked about my power supply, which was only 380W and he suggested that may not be enough juice for 4 HDD's with 6TB storage and could be the culprit.  I replaced with an Antec 500W, then yanked out and reformatted the bad drives - Windows formatter confirmed the bad sectors on one drive (and I did a successful RMA) but the other was clean, and still works without issue today.  At the time I asked if the bad sectors THEMSELVES could possibly have been "duplicated" onto another drive but Ken (or another moderator) thought this was impossible.

    But I failed to check the actual D: partition at the time.  My thought is:  This is the same issue, and has been out there since my low-voltage problems, but only now has surfaced as I've filled one hard drive to 98% and the D: drive is starting to be used.  But I suspect it's a PHANTOM issue!  I haven't done so yet (still waiting for the backup to my external 2TB drive to complete) but I bet once I take this drive out and reformat it, it will be fine.

    WHICH BRINGS ME, finally, to my latest question:  Rather than reformat the whole drive and do a painful Server Reinstall, is there ANY WAY to move the data off the D: partition to my other drives, reformat ONLY the D: partition, and leave the C: system volume (and WHS) intact?  I suspect the answer is no but have to ask.

    Thanks for listening!  Or reading, at least.

     

    Sunday, July 11, 2010 3:54 AM
  • So you're basically indicating that if a hard drive sector fails, the corrupted data on that sector is then copied to the duplicate shadow, making both the file and its duplicate useless.  You would think WHS would be smart enough to do a CRC check before duplicating the data!  It sounds like file duplication is only helpful in the event of a complete HDD failure, like a motor burnout - where the drive fails before the demigrator has time to update the duplicate file with the corrupted file.  Hmmph - sectors going bad is a much more common scenario, as old drives begins to deteriorate.  I feel the value of file duplication is suspect, and only not unreasonable b/c SATA HDD's are so cheap these days.

    So a nice, logical feature, if feasible, would be for the file duplicator to do that CRC check if it ever identifies a change in the file, recognize its now a corrupted file, and create a DIFFERENT original shadow file somewhere ELSE, using the uncorrupted DUPLICATE shadow.  Would that be something I could throw out as an enhancement request? 

    Sure, you can make that suggestion on Connect.  However, please note that Vail (the next version of WHS, now in beta) already has a CRC checksum as part of DE.  Since it's already part of Vail, it seems highly unlikely MS will go back to make a significant change to DE v1 at this point. 
    So back to my situation:  Chkdsk found 350MB of bad sectors of a 1.5TB drive - on the D: partition of my system drive.  HOWEVER, as it is a Seagate drive I was prompted to run the "Seatools" HDD diagnostic utility before RMA, which even after its "LDST" (Long Drive Self Test) found no problems whatsoever with the drive!  This makes me believe the drive itself is actually OK.  My reasoning (and sorry to hijack this thread, Verohomie) is this:  I had virtually the exact same size of bad sectors previously on 2 separate HDD's.  Ken Warren asked about my power supply, which was only 380W and he suggested that may not be enough juice for 4 HDD's with 6TB storage and could be the culprit.  I replaced with an Antec 500W, then yanked out and reformatted the bad drives - Windows formatter confirmed the bad sectors on one drive (and I did a successful RMA) but the other was clean, and still works without issue today.  At the time I asked if the bad sectors THEMSELVES could possibly have been "duplicated" onto another drive but Ken (or another moderator) thought this was impossible.

    But I failed to check the actual D: partition at the time.  My thought is:  This is the same issue, and has been out there since my low-voltage problems, but only now has surfaced as I've filled one hard drive to 98% and the D: drive is starting to be used.  But I suspect it's a PHANTOM issue!  I haven't done so yet (still waiting for the backup to my external 2TB drive to complete) but I bet once I take this drive out and reformat it, it will be fine.

    WHICH BRINGS ME, finally, to my latest question:  Rather than reformat the whole drive and do a painful Server Reinstall, is there ANY WAY to move the data off the D: partition to my other drives, reformat ONLY the D: partition, and leave the C: system volume (and WHS) intact?

    It would definitely not be supported.  I suppose if you shut down the server, pull the primary drive from it, connect it to another computer and copy everything from the D partition, format D, copy everything back to D, then put the drive back in your server, then it might work.  But, as I said, it's definitely not supported so if you decide to go that route, you might want to make sure you have backups of everything first. 

    I suspect the answer is no but have to ask.

    Thanks for listening!  Or reading, at least.

     

    Sunday, July 11, 2010 2:25 PM
    Moderator
  • So you're basically indicating that if a hard drive sector fails, the corrupted data on that sector is then copied to the duplicate shadow, making both the file and its duplicate useless.
    ...

    Not likely to be the case. Hard drives have error correction and detection, which will normally detect that a sector is having issues and try to A) recover good data from it and B) move that data elsewhere. So a bad sector won't spread to other disks. What may happen is that the bad sector may be determined to be "unrecoverable". In that case, you lose the data on that sector. Because of the way duplication works, there's a second copy of that sector on another drive, but (again because of the way duplication works) it's not going to be automatically "promoted" so you may have file access issues. Running chkdsk will usually let Drive Extender deal with the situation, so if you experience a "file conflict" error, or a report of an unhealthy disk in the console, your best course of action is chkdsk with the /r parameter.

    As for direct manipulation of the D: partition, it's possible but unsupported. The process would involve cloning the entire drive onto a new drive.


    I'm not on the WHS team, I just post a lot. :)
    Sunday, July 11, 2010 3:56 PM
    Moderator
  • But at the point I received the error, it sounded like the shadow file was gone or inaccessible.  Are you suggesting Ken that running chkdsk /r at that point might save the data?

    On my other project:  Confirmed, the drive is testing 100% stable, no errors, no bad sectors!  Very odd but a happy ending, so I will now go about the arduous task of rebuilding the server using "Server Reinstall" and I'm still going to use this same drive.

    Windows has checked the file system and found no problems.

    1465136127 KB total disk space.
       3093600 KB in 7 files.
            12 KB in 13 indexes.
             0 KB in bad sectors.
        111179 KB in use by the system.
         65536 KB occupied by the log file.
    1461931336 KB available on disk.
     

    Sunday, July 11, 2010 5:36 PM
  • Well...as it turns out just a day later I checked again and Windows marked almost 400GB of bad sectors.  Very odd...doing an RMA now, pulled an unused 500GB HDD out of storage for the system drive and had to server reinstall yet again.  That will teach me!

     

    Wednesday, July 14, 2010 3:41 PM
  • You can use the Acronis Drive Monitor (Free!) that Acronis recently made available.
    http://www.acronis.com/homecomputing/products/drive-monitor/

    I like it :-)


    .
    .
    .
    .

    "Bardmaster" wrote in message news:f098cf04-f29b-4825-a883-f1bed5441805@communitybridge.codeplex.com...

    Well...as it turns out just a day later I checked again and Windows marked almost 400GB of bad sectors. Very odd...doing an RMA now, pulled an unused 500GB HDD out of storage for the system drive and had to server reinstall yet again. That will teach me!


    Have a nice day!
    Wednesday, July 14, 2010 5:11 PM
  • Thanks, installed and trying it out!  Assume need to use AnyServiceInstaller to set it up as a service when no users logged in.
    Thursday, July 15, 2010 5:49 AM