none
100% CPU during backups RRS feed

  • Question

  • We run DPM 2010 (incl hotfix rollup 2250444) on windows 2008 SP2 64-bit
    This a VM on ESXi
    We granted 2 CPU’s (Xeon) @ 3 GHz and 12 GB memory

    The problem we experience is that during backups (hourly on disk, daily on tape) sqlserve.exe,  DPMRA.exe and WmiPrvSE.exe take all the cpu capacity.

    Most of the time it is only sqlserv.exe that takes 98% but is also happens that they cpu capacity is spread out between the 3 processes (sqlserve.exe,  DPMRA.exe and WmiPrvSE.exe)

    The result of this is that backups take much longer than they should. E.g a tape backup of 11GB takes an hour to complete where this would normally take less than 5 minutes.

    We didn’t always have this issue but it started somewhere along the way but I don’t know when exactly. I tried the rollup 2250444 but without a positive result.

    I’m also under the impression that this issue is stronger when we don’t close the DPM admin console. Most of the time me and my colleague leave it open in a disconnected RDP session.

    More people experience this problem?

    Any idea on how to troubleshoot this?

    Thx!

    Thursday, January 20, 2011 10:23 AM

Answers

  • 2 weeks ago I created a backup of the DPM database. Reinstalled the server and restored the DPM database and attached the iSCSI volumes again to the new server.

    Things are much more stable now. No high CPU usage for long periods of time.

    Now a clean install of win 2008 R2 x64 and DPM 2010

    Previous server had a long track record.

    Physiclal server with DPM 2007 > Upgraded to DPM 2010 RC > P2V ESXi > Upgraded to DPM 2010 RTM but because of some upgrade issues there were still some DPM 2007 databases left

     

    Friday, May 6, 2011 1:31 PM

All replies

  • Nobody?
    Friday, February 4, 2011 5:58 PM
  • Hi Mister,

     

    How many DS's are being protected by this DPM server. How many parallel Tape backups are you running? How many number of drives you have on the Library?

    Regards

    Mukul Shekhawat [MSFT]

    This posting is provided AS IS, with no warranties, and confers no rights.

    Monday, February 7, 2011 8:31 PM
  • Hi Mukul,

    Thanks for your reply!

    We have 5 different protection groups

    Clients (laptops) with 117 members. hourly backups to disk daily full backup to tape
    Extra with 2 members, monthly full backup to tape
    Production with 153 members. Every 2 hours to disk, daily full backup to tape
    VM's with 2 members, monthly full backup to tape
    Workgroup with 1 member and a weekly full backup to disk

    Total of 275 members that are protected

    The tape unit is a Dell TL2000 tape library with 1 LTO4 tape drive and a iSCSI interface.
    All tape backups are scheduled so they don't run at the same time.

    Regards,
    Bas

     

     

    Wednesday, February 9, 2011 10:45 AM
  • Hi,

    Have you noticed if this problem can be reproduced with only client jobs running or do you have the same problem if other jobs are running, but not client protection jobs?

    Thanks,

    Marc

    Wednesday, February 9, 2011 1:56 PM
    Moderator
  • Hi,

    Have you noticed if this problem can be reproduced with only client jobs running or do you have the same problem if other jobs are running, but not client protection jobs?

    Thanks,

    Marc


    Mabey it is indeed only happening when client synchronisation jobs are running.
    DPM was doing a tape back up and cpu usage was normal/low. But then 3 client synchronisation jobs started and CPU usage of sqlservr.exe went to 50%

    Is there a know issue with client synchronisation and cpu usage?

    Friday, March 4, 2011 10:36 AM
  • Mr. Iks,

    We recently moved three servers to ESXi 4.1 and noticed that when I did a Replica Synchronization during working hours the AP server went to nearly 100% and was causing an issue with our ERP App.

    I throttled that server back during working hours to 2056 kbps and now it doesn't interfer with their work, although it does take longer to run.  I couldn't wait until after hours as we are EST and have peiople connecting PST so they are around 3 hour longer than me.  Besides I kept forgetting to start it.

    DPM is 2007 on a W2K3 32-bit server.


    Peter Jam
    Thursday, April 14, 2011 7:16 PM
  • Hello,

    If you have compression on the wire enabled then it will consume some of the CPU cycles on both the DPM and target server.  I've never seen it peg out the CPU to 100% though.  If you are having the DPM server perform an eseutil on a Exchange storage group, this will also cause a CPU spike as well.


    Thanks,
    Shane
    Friday, May 6, 2011 1:26 PM
  • 2 weeks ago I created a backup of the DPM database. Reinstalled the server and restored the DPM database and attached the iSCSI volumes again to the new server.

    Things are much more stable now. No high CPU usage for long periods of time.

    Now a clean install of win 2008 R2 x64 and DPM 2010

    Previous server had a long track record.

    Physiclal server with DPM 2007 > Upgraded to DPM 2010 RC > P2V ESXi > Upgraded to DPM 2010 RTM but because of some upgrade issues there were still some DPM 2007 databases left

     

    Friday, May 6, 2011 1:31 PM
  • Great that it works for you. In the feature look at applying different prefomance counters to determine bottle-necks in your solution.

     

     http://technet.microsoft.com/en-us/library/ff399366.aspx


    Best Regards

    Robert Hedblom

    MVP DPM

     


    Check out my DPM blog @ http://robertanddpm.blogspot.com


    Friday, May 6, 2011 8:23 PM
    Moderator