none
WHS rebuild questions RRS feed

  • Question

  • Looking for advise, opinions, and input!

    What is the "correct" procedure to return a rebuilt server to its previous working state.

    WHS rig w/ (2) 1.5T drives...  The drive w/ the OS crashed.
    Photog business needs continued, uninterrupted R/W access to the images from that server.
    Removed 2nd drive from server rig and placed in workstation as current work-around.

    Now, to rebuild the server, I have to start by replacing that dead drive.

    Scenario 1:  Rebuild the rig as it was before the drive crash, replacing only the (1) 1.5T drive.
    What is the procedure for this?   Here's how I think it works:

      A) Place new drive in rig as only HD.
      B) Reinstall WHS using same workgroup name, server name, password
      C) Reinstall into the rig the 2nd drive that still has all the old shares and other old server files on it 
      D) Turn on the WHS and it magically recognizes that new drive needs to have all the files from old drive copied onto itself so duplication begins automatically and the server is ready for business again the next AM.

    Is that right? If not, what is correct process?

    But wait!  Maybe I can make lemonade... This may be a good opportunity to enbiggen the storage pool!

    Scenario 2:  Build a BIGGER server! 
    They really do need a LOT of storage space so perhaps I should just build a BIGGER server! 
    2T drives seem to be a sweet spot. 
    Should I get 2, 3, 4, or more drives?

    As far as WHS goes, what is the optimal number os same-size drives?

    Under this 'build it bigger' scenario I assume it'd be easiest to just build the server anew (with whatever is the 'right' number of drives) THEN COPY (from the old 2nd drive, now in the  workstation) all the old shares and other old files onto it.

    For bonus points:

    Isn't there a SMART util thatll broadcast if a drive is going to fail?

    Thx all!

    Friday, November 5, 2010 8:13 AM

Answers

  • Scenario 2 is your best bet. Use two of the largest drives you can afford, use duplication for everything, and make sure you can take regular backups off-site, so pick up an additional external drive (or two, so you can rotate them off-site) that's large enough to hold your data.

    As for S.M.A.R.T.: Google is one of the world's largest consumers of hard drives. A few years ago they looked at hard drive failures, and published a research paper on the subject. They drew several interesting conclusions, among them:

    • There's a low correlation between drive temps and failure rates.
    • There's a low correlation between drive activity and failure rates.
    • The presence of certain S.M.A.R.T. indicators is a good indicator that the drive is about to fail. There are four that are called out as particularly telling; they're of the "if you see this pop up, replace your drive ASAP" variety in my opinion.
    • The lack of S.M.A.R.T. indicators is not a good indicator that the drive is healthy. A large fraction of drive failures were unannounced by S.M.A.R.T.

    The original paper is here.


    I'm not on the WHS team, I just post a lot. :)
    • Marked as answer by tech lab Friday, November 5, 2010 6:33 PM
    Friday, November 5, 2010 1:29 PM
    Moderator

All replies

  • There will not be magical detection of an old drive.

    If you select a Server reinstall (not a new install!), the functioning old drive has to be connected already, and if you are lucky, it will be detected and the data integrated. User accounts and passwords are lost with the OS.

    Other ways are described in the FAQ How to recover data after server failure.

    The problem with SMART tools is, that even if they note a failure in time, not each disk fault happens announced and is predictable. So it would only cover some specific situations (or eventually flood you with messages even for not so important parameters)

    Also the entire server can go down due to overvoltage, fire, theft, user error ...

    Therefore regular backup to external media and taking this media offline is mandatory for files, which are considered to be important. The ability to do this should be part of any server dimensioning.

    Best greetings from Germany
    Olaf

    Friday, November 5, 2010 10:43 AM
    Moderator
  • Scenario 2 is your best bet. Use two of the largest drives you can afford, use duplication for everything, and make sure you can take regular backups off-site, so pick up an additional external drive (or two, so you can rotate them off-site) that's large enough to hold your data.

    As for S.M.A.R.T.: Google is one of the world's largest consumers of hard drives. A few years ago they looked at hard drive failures, and published a research paper on the subject. They drew several interesting conclusions, among them:

    • There's a low correlation between drive temps and failure rates.
    • There's a low correlation between drive activity and failure rates.
    • The presence of certain S.M.A.R.T. indicators is a good indicator that the drive is about to fail. There are four that are called out as particularly telling; they're of the "if you see this pop up, replace your drive ASAP" variety in my opinion.
    • The lack of S.M.A.R.T. indicators is not a good indicator that the drive is healthy. A large fraction of drive failures were unannounced by S.M.A.R.T.

    The original paper is here.


    I'm not on the WHS team, I just post a lot. :)
    • Marked as answer by tech lab Friday, November 5, 2010 6:33 PM
    Friday, November 5, 2010 1:29 PM
    Moderator
  • ... select Server reinstall (not new install!) ... and if you are lucky, it will be detected and the data integrated...
    Humm...  I think I'll just pop in some bigger disks, do a new install, then copy of the old data from the remaining good drive!

    Other ways are described in the FAQ How to recover data after server failure .

    Thanks for that.
    Friday, November 5, 2010 6:30 PM
  • Thanks for another excellent answer.

    You do suggest above to use TWO drives but I have seen elsewhere that it may be best to start with ONE (the system drive) then add drives in pairs meaning the optimal starting point for a machine in production would be THREE drives.

    Thanks again, Ken!

    and fascinating paper BTW.

    Friday, November 5, 2010 6:41 PM
  • You do suggest above to use TWO drives but I have seen elsewhere that it may be best to start with ONE (the system drive) then add drives in pairs meaning the optimal starting point for a machine in production would be THREE drives.

    Thanks again, Ken!

    and fascinating paper BTW.

    At one time I'd have said the same thing; when the "sweet spot" in the $/GB equation was well under a terabyte, it made sense to use a smaller drive as your system drive, and then add storage drives in pairs. Today, not so much. You can buy a 2 TB drive for $100 plus shipping right now.

    Buy three, two bare and one in an external enclosure (for backups).


    I'm not on the WHS team, I just post a lot. :)
    Friday, November 5, 2010 11:53 PM
    Moderator