Results 1 to 15 of 15
  1. #1
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts

    A few more RAID server drive questions

    Just want to check my theory about drives in a virtual drive group.
    Suppose you have 3 drives arranged in a RAID 5 configuration, all part of the same virtual drive, with no dedicated hot spare.
    And suppose one drive craps out and you need to replace it.
    If I understand correctly, as soon as you physically pull the drive and swap in a new one, PERC 6 will call that a foreign configuration.
    Then you will have to open the configuration utility, and import the new configuration, whereupon the controller will rebuild the virtual drive as before, using the replaced HDD?
    Or do you have to "rebuild" the new drive first?

    Also, SAS drives can be hot swapped, but SATA drives cannot, correct?
    I understand SAS and SATA drives cannot be used in the same VD, but can be mixed on the same controller as different VD's?

    Thanks,
    rstew

  2. #2
    WS Lounge VIP
    Join Date
    Dec 2009
    Location
    Earth
    Posts
    8,172
    Thanks
    47
    Thanked 981 Times in 911 Posts
    Server RAID controllers usually allow hot swap and automated re-builds. The PERC series are no exception, but some drives are not hot swappable (don't use them in a production environment, home use is fine).
    The basic arrangement is:
    1. Run RAID management software and determine what is wrong.
    2. If the faulty drive is marked as off-line and hot swappable, pull it out for 30 seconds and then put it back. 50% of the time the drive will come back to life and the array will automatically re-build.
    3. If the drive has gone to god, replace it with a new blank drive and the array will automatically re-build.
    4. If you use a drive with existing data the array will consider that a foreign disk and await your instructions. Do not import the configuration, you will break the array and lose your data. Mark the drive as available (I can't remember the correct term) and the array will use the new drive and re-build.

    If you need to move an existing array of disks to a new host, 'cause the old one died, you put them in the new host and import the "foreign" configuration to retain the array and data.

    Do not mix disk interfaces on a single controller. If it goes pear shaped....

    cheers, Paul

  3. #3
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Paul;
    OK thanks; all good to know.
    Not sure what you mean by: "Do not mix disk interfaces on a single controller."
    By interface do you mean different drive types, like SAS and SATA? Or?
    The PERC 6 manual says it can support up to 32 virtual drives per controller, so that's not it.
    Can you provide more details?

    thanks,
    rstew

  4. #4
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Another discovery:
    Just to test another older PERC 5 2950 system (I got for parts), I swapped in two PERC 6 drives from the 1950, without thinking about issues the different PERC revs may cause.
    It booted up fine and loaded the OS without issue; so PERC 6 is backwards compatible.
    Then when I popped the drives back into the PERC 6 system, it came up with a foreign array warning. Oh-oh!
    Went into the RAID utility and imported the array, and all was well.
    Lesson learned I guess.

    rstew
    Last edited by rstew; 2015-10-30 at 12:54.

  5. #5
    WS Lounge VIP
    Join Date
    Dec 2009
    Location
    Earth
    Posts
    8,172
    Thanks
    47
    Thanked 981 Times in 911 Posts
    Yes, SATA and SAS together. You must have only one type per array, not mixed, but you can have both on the controller.

    Swapping disk sets between machines sometimes requires importing, sometimes not. No big issue.

    cheers, Paul

  6. #6
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Good info Paul. Thanks, I am starting to see my way along. All fun so far.

    On reading through the Perc 6 manual further, I see it supports Raid Level Migration, as well as expansion of existing arrays, though there are no specific instruction on how to do either. Nor does either of these options appear in the setup utility as far as I can see.
    I guess I will have to play around some more.

    rstew

  7. #7
    WS Lounge VIP
    Join Date
    Dec 2009
    Location
    Earth
    Posts
    8,172
    Thanks
    47
    Thanked 981 Times in 911 Posts
    Expansion is just a matter of adding a new disk to the array, the rest is automatic.
    I've never migrated RAID and can't see a reason why you'd do it, but it makes sense to add while you're adding features.

    cheers, Paul

  8. #8
    4 Star Lounger
    Join Date
    Jan 2010
    Location
    Fort McMurray, Alberta, Canada
    Posts
    557
    Thanks
    51
    Thanked 68 Times in 66 Posts
    When you say "RAID 5 configuration, [...], with no dedicated hot spare", I'm not sure what you mean.

    RAID configurations are precisely specified. By definition RAID 5 contains one parity drive and thus an array can withstand the failure of a single drive in a RAID pool. However running the RAID array without the parity drive is considered an impaired configuration and is merely a stopgap until the failed drive can be replaced. The failure of a second drive from an impaired configuration is fatal to the entire array.

    Re: "Or do you have to 'rebuild' the new drive first?" No, this is wrong and your preceding statement was correct. There's no way to rebuild the drive without adding it back to the array. Even if you have backups, and that includes bit-image perfect image backups, they are useless for this purpose. RAID systems calculate parity constantly and even the tiniest amount of activity on the RAID array would cause divergence from the image backup. Thus no restore from backup would be reliable.

    You probably know this already but returning a RAID array to it's protected state after a drive failure, involves reading every drive in the array and recalculating the parity bits. Thus this activity takes several hours and involves enormous amounts of I/O. The array is usable during this time but it's performance is degraded.

    As for the specifics of the hot-swapping capabilities, I don't know. Typically you want to get specific advice from the vendor involved, whether from a manual, online, phoning tech support or whatnot. Your situation is atypical but in a production environment, you simply cannot afford to assume that hot-swapping is supported and then crash the system because it wasn't.

  9. #9
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Quote Originally Posted by Paul T View Post
    Expansion is just a matter of adding a new disk to the array, the rest is automatic.
    I've never migrated RAID and can't see a reason why you'd do it, but it makes sense to add while you're adding features.

    cheers, Paul
    I currently have 3 drives in a RAID 5 configuration, and I was thinking to add a 4th drive and migrate to a RAID 6 configuration, really just to see if it would work.
    Thus the questions. Looks like PERC 6 can do that; I just have to play with it and figure out how. As you say the expansion should be automatic hopefully.

    rstew

  10. #10
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Quote Originally Posted by BHarder View Post
    When you say "RAID 5 configuration, [...], with no dedicated hot spare", I'm not sure what you mean.

    RAID configurations are precisely specified. By definition RAID 5 contains one parity drive and thus an array can withstand the failure of a single drive in a RAID pool. However running the RAID array without the parity drive is considered an impaired configuration and is merely a stopgap until the failed drive can be replaced. The failure of a second drive from an impaired configuration is fatal to the entire array.

    Re: "Or do you have to 'rebuild' the new drive first?" No, this is wrong and your preceding statement was correct. There's no way to rebuild the drive without adding it back to the array. Even if you have backups, and that includes bit-image perfect image backups, they are useless for this purpose. RAID systems calculate parity constantly and even the tiniest amount of activity on the RAID array would cause divergence from the image backup. Thus no restore from backup would be reliable.

    You probably know this already but returning a RAID array to it's protected state after a drive failure, involves reading every drive in the array and recalculating the parity bits. Thus this activity takes several hours and involves enormous amounts of I/O. The array is usable during this time but it's performance is degraded.

    As for the specifics of the hot-swapping capabilities, I don't know. Typically you want to get specific advice from the vendor involved, whether from a manual, online, phoning tech support or whatnot. Your situation is atypical but in a production environment, you simply cannot afford to assume that hot-swapping is supported and then crash the system because it wasn't.
    BHarder;
    Thanks for your insights.
    Many of the things I have been asking are borne of just not knowing exactly how all this stuff works, so just bear with me.
    PERC 6 allows you to designate a hot spare with a virtual array apparently, which would allow automatic switchover and rebuilding in the case of one of the main drives failing.
    I do understand that a RAID 5 array requires a minimum of 3 drives for redundancy, and that if one fails you are on borrowed time.
    I have come to understand that drive rebuilding is done automatically once the failed drive in an array is replaced. And I surely did see that it took a huge amount of time when it did happen!
    I also read that if you have 4 drives, it is better to set up a RAID 6 configuration using all 4, than have a 3 drive RAID 5 configuration with a hot spare. I am not sure everyone would agree with that though?

    Anyway, thanks to yourself and PaulT for helping me along the road to understanding!
    rstew

  11. #11
    WS Lounge VIP
    Join Date
    Dec 2009
    Location
    Earth
    Posts
    8,172
    Thanks
    47
    Thanked 981 Times in 911 Posts
    Quote Originally Posted by rstew View Post
    I also read that if you have 4 drives, it is better to set up a RAID 6 configuration using all 4, than have a 3 drive RAID 5 configuration with a hot spare
    RAID 6 is 5 with an additional parity disk, so when you lose a disk you still have RAID 5 protection. Unless you are running a 24/7 production system you do not need RAID 6 - you also lose capacity, n-2 x disk capacity.

    cheers, Paul

  12. #12
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Quote Originally Posted by Paul T View Post
    RAID 6 is 5 with an additional parity disk, so when you lose a disk you still have RAID 5 protection. Unless you are running a 24/7 production system you do not need RAID 6 - you also lose capacity, n-2 x disk capacity.

    cheers, Paul
    Ok I get it Paul. Sounds like overkill for my modest needs.
    i'll just stay with Raid 5 as you suggested all along.

    thanks
    rstew

  13. #13
    4 Star Lounger
    Join Date
    Jan 2010
    Location
    Fort McMurray, Alberta, Canada
    Posts
    557
    Thanks
    51
    Thanked 68 Times in 66 Posts
    Re: "PERC 6 allows you to designate a hot spare..."

    OK, interesting. It sounds like the vendor has added a proprietary feature on top of standard RAID functionality there. Just a comment and a point of interest though, this sounds very much like a RAID 6 configuration, particularly as the spare is hot. A hot spare implies that drive is running (I think...), so maybe the PERC isn't doing any I/O on that hot spare? This is just speculation on my part.

    I'd be more interested in a cold spare, where the additional drive is completely dark until needed. And quite honestly you can simulate that by buying and mounting a drive and simply not hooking up the power and data. Either that or go full-on RAID 6.

    However these are enterprise considerations for high availability systems. RAID 5 is popular because it gives you decent redundancy against the most common failure types. And it doesn't so much cost that the cost becomes a problem.

  14. #14
    3 Star Lounger
    Join Date
    Jun 2014
    Posts
    283
    Thanks
    11
    Thanked 6 Times in 5 Posts
    Yes I agree a "hot spare" likely means it is spinning and ready to take over if the controller detects that one of the main array drives has cratered.
    According to the manual this all happens seamlessly and when the dead drive is physically replaced the hot spare automatically reverts back to a hot spare. Kind of a nice feature.
    This is all covered in the PERC 6 User Manual which you can download from the Dell site. It is "26-2923.PERC6UsersGuide" and its a .pdf format.
    Hot spare is listed as a feature on page 23, and the actual setup procedure is covered on page 94.

    This is called a "dedicated hot spare" (dedicated to the array it is attached to) and then Perc 6 also provides for setting up a "global spare", which can replace a failed drive in any array on the controller it is attached to. You are limited only by the number of drive bays you have available.
    I like your idea of a cold spare, which I guess you would simply swap into the spot where the failed drive is shown (by an amber drive LED). Then the controller would simply and automatically rebuild the drive, as PaulT has said.

    But again for a simple home system such as what I am contemplating, a simple 3 drive RAID 5 setup with a swappable cold spare should work just fine.
    I am using standard Seagate 1 TB SATA 7200 drives, so I would need to shutdown to do the swap, but no big deal.

    Anyway thanks for the insights. We are making progress!
    rstew

  15. #15
    WS Lounge VIP
    Join Date
    Dec 2009
    Location
    Earth
    Posts
    8,172
    Thanks
    47
    Thanked 981 Times in 911 Posts
    Hot spares are standard in enterprise systems - set up plenty of them. If the hot spare fails on its own you just replace it as you would any other disk.

    cheers, Paul

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •