Cognos Analytics

Cognos Analytics

Connect, learn, and share with thousands of IBM Cognos Analytics users! 

 View Only

  • 1.  Schedules failing DPR-ERR-2072 Unable to load balance a request with absolute affinity, since upgrade to 11.1.5

    Posted Mon February 03, 2020 09:31 AM
      |   view attached
    Hi community,

    Since we upgraded one of our clients to 11.1.5, they have having an issue with schedules failing with the following error:

    DPR-ERR-2072 Unable to load balance a request with absolute affinity, most likely due to a failure to connect to the remote dispatcher. See the remote dispatcher detailed logs for more information. Check the health status of the installed system by using the dispatcher diagnostics URIs. CAF-WRN-2082 An error has occurred. Please contact your administrator. The complete error has been logged by CAF with SecureErrorID:2020-01-30-06:00:25.135-#466

    • As mentioned, they has only been since upgrading to 11.1.5 from 11.0.13, it has been working for several months.
    • It is always seemingly happening at 6:00am, when a lot of schedules are being kicked off (80).
    • We have attempted to tweak some of the settings to the batch service, such as batch report processes from 2 to 4; peak time to end at 8am; report queue timeout from 240 seconds to 280 seconds.
    • Schedules are running fine before, between (in some cases), and after. The same reports that fail are running fine previous days also.


    Does anybody have any suggestions? We are probably going to try a reinstall but wondered if there other tuning changes that could resolve the above. 

    Thanks,

    Ashley

    ------------------------------
    Ashley Kent
    ------------------------------

    #CognosAnalyticswithWatson


  • 2.  RE: Schedules failing DPR-ERR-2072 Unable to load balance a request with absolute affinity, since upgrade to 11.1.5

    Posted Tue February 04, 2020 06:36 AM
    ​Hi Ashley

    Our experiences in solving this problem are based on CA 11.1.3 and we haven't tested this on 11.1.5 yet.

    There are two aspects at play here:  1) the tuning of the server according to the hardware and  2) Buffering of the reports you have scheduled to run at the same time.

    Even when we tuned the server for normal concurrent users, running the scheduled reports could still cause this error.  Check that you have tuned your system properly first, according to the guide:
    https://www.ibm.com/support/knowledgecenter/en/SSEP7J_11.1.0/com.ibm.swg.ba.cognos.ug_cra.doc/c_setthemaximumnumberofprocessesandconnections.html#SettheMaximumNumberofProcessesandConnections

    We were able to generate these errors by overloading the server with nested schedules, that would execute concurrently.  The resolution was to change the shape of the load, that we are putting on the server at any single moment in time.  We feed the reports gradually to be produced instead of running them all at once.

    This error seems random because the load is not exactly the same every time.  The error thus happens when it enters the overloaded state and try to recover from it.  That is why the problem is not with the actual reports.  It could be that you have to restart the CA service to recover after this error event, to ensure a reliable recovery.

    I assume you are checking that there are no deadlocks on the content store database.

    Also remember that the burst reports are by definition more server intensive reports than the normal single reports.  We don't like to run burst reports concurrently, while the single reports are much easier to run concurrently.  As a convention, we reserve around half the server resources for the burst activity and then load single reports according to the other half of the capacity. 

    Test what your system can realistically handle, without producing these errors and then stay below that threshold.

    Hope it helps.

    ------------------------------
    Jan Vermeer
    ------------------------------

    Original Message


  • 3.  RE: Schedules failing DPR-ERR-2072 Unable to load balance a request with absolute affinity, since upgrade to 11.1.5

    Posted Tue February 04, 2020 08:19 AM
    Hi Ashley,

    Our company has had the same issue ever since upgrading from 11.0.13 to 11.1.3 .   Spreading the schedules apart a few minutes does help and now we only get the issue once in a while, usually around month ends. However, I have also questioned the community as to why the same server that did not have issues running our schedule on 11.0.13  all of a sudden struggles when on 11.1.x. Please post if you find a better answer than "tune your server".

    Best of Luck,

    ------------------------------
    Nicholas Youndt
    ------------------------------

    Original Message


  • 4.  RE: Schedules failing DPR-ERR-2072 Unable to load balance a request with absolute affinity, since upgrade to 11.1.5

    Posted Wed February 05, 2020 12:55 PM
    Hi Ashley - my take on this is that there is good advice on spreading out the load, but bottom line is that with each incremental release of additional features on this platform, you should expect additional system overhead.  In most environments there is typically plenty of idle capacity, but we need to be more cautious about over the top upgrades because even minimal increases in application footprint can overload certain environments.
    Also, be sure you are tuning the relevant dispatchers if you are running a distributed environment.

    ------------------------------
    Albert Valdez
    albert@intelalytics.com
    ------------------------------

    Original Message


  • 5.  RE: Schedules failing DPR-ERR-2072 Unable to load balance a request with absolute affinity, since upgrade to 11.1.5

    Posted Thu February 06, 2020 09:18 AM
    Hi Ashley. Are these CQM or DQM reports?

    ------------------------------
    TREVOR COMEAU
    ------------------------------

    Original Message


Related Content