none
DPM 2012 R2 Slow Consistency Checks on File Servers RRS feed

  • Question

  • Hey All, has anyone noticed just how slow consistency checks are for larger volumes? I know that this seems to be a grin and bear it currently but it's a bit of a showstopper having to miss backups due to it.

    In our case a volume on one of our file servers contains about 1.8TB of data (not exactly insanely huge) and the other night it bluescreened (now sorted) just before the back which then triggered DPM to start a consistency check, this took over 22hrs to complete which meant we missed the backup window completely and had no option but to wait it out. I've checked the usual bottlenecks (see link below for more) and we pretty much as good as we're going to get for speed on that side for connectivity server resource and disk speed as nothing is stretched at either end and we get good transfer rates.

    I've been looking at ways to mitigate against it and we currently have one recovery point daily with multiple syncs (will syncs help without without a recovery point?) and have just been looking at and old reply from Mike Jacquet on this thread and there was a recommendation where possible to start using volume mount points as a workaround. Is this still one of the only options? We'd rather not have to undertake that level of work and cost on our file servers if we can help it.

    Any info or experience of optimising CC's would be greatly appreciated as this kind of thing could seriously effect SLA's etc.

    Sunday, November 16, 2014 11:43 PM

Answers

  • Hi,

    The frequency of synchronizations should not be a factor except it does increase the likelihood that one may fail in such a way that it marks the replica inconsistent.   I always think of Synchronization frequency as sipping thru a straw and bring over smaller amount of changes throughout the day, or drinking from a fire hose and have the changes all come over at the end of the day before the RP is made.   More frequent synchronizations does protect you better since more changed file data is applied to the replica, and is available for recovery after making a manual recovery point.  So it's really up to your SLA requirements on how you handle the synchronization frequency.


    Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    • Marked as answer by [Malco] Tuesday, December 9, 2014 8:36 PM
    Tuesday, December 9, 2014 5:48 PM
    Moderator

All replies

  • Hi,

    We are currently investigating how we can either speed up consistency checks, or be able to maintain recovery point schedule while CC is ongoing.  In the meantime, best solution would be to virtualize the workload and back it up as a Hyper-V guest.  CC would the be against the larger .VHD file and not million of smaller files.


    Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    Tuesday, December 2, 2014 11:21 PM
    Moderator
  • Hi,

    Can you share some details about the protected volume in term of the number of files (run a read only chkdsk) and perhaps some CC job details that show number of items scanned / fixed, elapsed time ?


    Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    Wednesday, December 3, 2014 12:25 AM
    Moderator
  • Hi Mike, sorry about the delay getting back to you. Hyper-V is not an option for us I'm afraid but I can see your point.

    We are currently replacing some of our file servers so are going gown the separate volumes/mount point route to spread the data out a bit and help us with a few other things too. This will effectively split the data out into 3 volumes in this case which will help us greatly. In the meantime do more frequent sync's help us? We will still only be taking one RP a day.

    Tuesday, December 9, 2014 2:39 AM
  • Hi,

    The frequency of synchronizations should not be a factor except it does increase the likelihood that one may fail in such a way that it marks the replica inconsistent.   I always think of Synchronization frequency as sipping thru a straw and bring over smaller amount of changes throughout the day, or drinking from a fire hose and have the changes all come over at the end of the day before the RP is made.   More frequent synchronizations does protect you better since more changed file data is applied to the replica, and is available for recovery after making a manual recovery point.  So it's really up to your SLA requirements on how you handle the synchronization frequency.


    Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    • Marked as answer by [Malco] Tuesday, December 9, 2014 8:36 PM
    Tuesday, December 9, 2014 5:48 PM
    Moderator
  • Good to know. Thanks for the info Mike, much appreciated as details surrounding this sort of stuff are pretty thin on the ground. 
    Tuesday, December 9, 2014 8:36 PM
  • Hi Malco,

    We are working on improving DPM consistency check performance for file servers with huge number of files.

    Can you please share some details about the protected volume in term of the number of files (run a read only chkdsk) and perhaps some CC job details that show number of items scanned / fixed, elapsed time ?

    This above information will be really useful to understand the issue and provide a fix in DPM.

    Tuesday, January 6, 2015 6:49 AM
    Moderator
  • We have a similar issue with a deduplicated volume. Taking ages to do a consistency check, > 12 hours.

    Details fro Get-DedupStatus:

    FreeSpace    SavedSpace   OptimizedFiles     InPolicyFiles      Volume
    ---------    ----------   --------------     -------------      ------
    1.04 TB      1.99 TB      765483             765344             E:

    Thanks,

    Marcus

    Wednesday, February 11, 2015 5:36 PM