locked
Windows/SQL 2008R2 - Logman & Perfmon are very slow RRS feed

  • Question

  • Hi, I have this box (Specs below) only for performance capture other SQL instances and directing output to SQL database on the same server. I have 22 data collector sets running on the box for SQL instances and with many counters for each instance .. with recording interval of one minute.

    I always find it extreemely slow to access these data collector sets .. for example .. "logman query" command from the command line is taking few minutes just to respond and list the running data collector sets. Also, when checking the counter values in the database, many times I found gaps of few minutes with no values recorded.

    Do you have any feedback on this situation and how to make it better.

    Regards,,

    System Manufacturer: 'HP', System Model: 'ProLiant DL580 G5'.
    Intel(R) Xeon(R) CPU E7340 @ 2.40GHz
    Logical CPU Count: 16 Hyperthread Ratio: 4 Physical CPU Count: 4
    Windows 2008R2 Enterprise 64bit
    SQL 2008R2 Enterprise 64bit
    52GB physical Memory - SQL is running with 20G

    Monday, July 16, 2012 9:44 PM

Answers

  • Hi Salem,

    We faced the same issue in so many servers which are hosted by HP. The perfmon used to be terribly slow and Hang after we installed a patch. Then we informed HP the same followed up with the Hardware they invovled MSFT and MSFT asked them to upgrade the Boot Files and some drivers and after that everything runs perfect for us.

    So i think you need to follow up with your Hardware Vendor regarding this.

    Regards..

    Hemanth

    Vote as Helpful if you find my answer useful..

    • Marked as answer by Maggie Luo Thursday, August 9, 2012 3:29 AM
    Wednesday, July 18, 2012 11:29 AM
  • How many pointers are you collecting?  Maybe it is taking while to read all of them?

    I personally have not configured such configuration, so I will not be able to provide you much more feed back sorry :(.  I usually rely on SCOM for my KPI collection from multiple instances.


    Mohit K. Gupta --- MCTS, MCITP, MCC2011
    My Blog OpsVault My MSDN Blog Twitter @SQLCAN

    • Marked as answer by ymsalem Thursday, August 9, 2012 4:33 AM
    Wednesday, July 18, 2012 5:14 PM

All replies

  • Missing values usually indicate CPU pressure, your system was too busy to write collection out.  IMHO having 22 data collectors against a single server is overkill.  What are you trying to achieve?

    Mohit K. Gupta --- MCTS, MCITP, MCC2011
    My Blog OpsVault My MSDN Blog Twitter @SQLCAN

    Monday, July 16, 2012 11:41 PM
  • I dont see any CPU pressure on the box .. it is mostly idle .. this is the central database for storing Perfmon counter data and I have 22 data collectors for the 22 SQL instances we have in production.
    Tuesday, July 17, 2012 2:28 AM
  • So you are collecting from remote instances into a central location; in which case skip was most likely due to network.  The information could not be pulled over, do all the log files skip the same time frame?  The server where the information was skipped how was the CPU on those servers?

    Mohit K. Gupta --- MCTS, MCITP, MCC2011
    My Blog OpsVault My MSDN Blog Twitter @SQLCAN

    Tuesday, July 17, 2012 8:03 PM
  • Thanks for replies and the follow up. Yes .. 22 collecters collecting from 22 SQL instances on the same domain but different machines. Perfmon Data Collecters are running on this dedicated machine with one SQL instance hosting the Repository database for Perfmon counters from those 22 collectors. Nothing else is running on this dedicated machine.

    Yeah . . Network would make sense .. but hard to capture .. will try anyways. Hard to tell if they all failing at the same time .. too many counters to compare when each one is skipping values.

    Most of the monitored instances are usually with no CPU pressure at all.

    My other and main concern is the behavior of Perfmon and logman on this dedicated machine .. extremely slow for both .. example as I mentioned above .. "logman query" will take few minutes to come back with results .. although the machine does not show any CPU/Memory/IO pressure.

    Regards,,

    Tuesday, July 17, 2012 8:18 PM
  • Hi ymsalem,

    I think we can try to exclude network issue firstly, i would like to recommend a tool named FPing, please refer :

    http://www.kwakkelflap.com/fping.html

    if you any question on the tool please let me know.

    Regards,

    nicofer


    • Edited by nicofer Wednesday, July 18, 2012 6:03 AM
    Wednesday, July 18, 2012 5:59 AM
  • Hi Salem,

    We faced the same issue in so many servers which are hosted by HP. The perfmon used to be terribly slow and Hang after we installed a patch. Then we informed HP the same followed up with the Hardware they invovled MSFT and MSFT asked them to upgrade the Boot Files and some drivers and after that everything runs perfect for us.

    So i think you need to follow up with your Hardware Vendor regarding this.

    Regards..

    Hemanth

    Vote as Helpful if you find my answer useful..

    • Marked as answer by Maggie Luo Thursday, August 9, 2012 3:29 AM
    Wednesday, July 18, 2012 11:29 AM
  • How many pointers are you collecting?  Maybe it is taking while to read all of them?

    I personally have not configured such configuration, so I will not be able to provide you much more feed back sorry :(.  I usually rely on SCOM for my KPI collection from multiple instances.


    Mohit K. Gupta --- MCTS, MCITP, MCC2011
    My Blog OpsVault My MSDN Blog Twitter @SQLCAN

    • Marked as answer by ymsalem Thursday, August 9, 2012 4:33 AM
    Wednesday, July 18, 2012 5:14 PM