Automic Workload Automation

 View Only
  • 1.  Sizing of JWPs, DWPs, WPs + Session Timeouts

    Posted Jul 02, 2019 12:56 PM
    Edited by Michael TMO8Y2tl Jul 02, 2019 12:58 PM
    Dear community,

    I'm currently facing the issue that I got timeout issues on AWI.
    e.g. displaying executions works fine, whereas filtering for deactivated tasks (e.g. search for archive keys) does not work.
    Which processes are in charge of that?

    Sometimes the AWI is blocked for all users / They are getting timeouts.
    Within the MQ-Records I saw that in most of the cases  MQJWP has a value of >40.
    Are there any experience which value is in indicator that the process sizing is to low?

    MQ-Records
    MQPWP
    1
    0
    0
    0
    MQWP
    24
    0
    0
    0
    MQDWP
    0
    MQMEM
    328
    MQRWP
    0
    0
    0
    0
    MQOWP
    0
    0
    0
    0
    MQJWP
    0



    Furthermore I noticed that users does not disconnect automatically once they have logged in. On which area this setting can be adjusted? Within Tomcat or in some AWI configuration? Is there any data which need to be cleaned up from time to time in order to increase the AWI performance?

    The old java client was much more faster.



  • 2.  RE: Sizing of JWPs, DWPs, WPs + Session Timeouts

    Posted Jul 02, 2019 02:45 PM
    Hi

    how many concurrent users do you have?

    and how many WPs, DWPs and JWPs?

    cheers, Wolfgang

    ------------------------------
    I know I do really know it!
    ------------------------------



  • 3.  RE: Sizing of JWPs, DWPs, WPs + Session Timeouts

    Posted Jul 02, 2019 03:38 PM
    Edited by Michael TMO8Y2tl Jul 02, 2019 03:54 PM

    Hi Wolfgang,

    Actually, only ~30-40 concurrent User. But one user is logged on to several clients + sometimes a user has several connections for same client open.
    Hence about 70-80 connections. That's why I'm not sure how to handle the processes.

    Currently the setup looks like this:
    2 x 2 x CPs
    2 x 1 x JCP
    2 x 13 x DWPs
    2 x 4 WPs
    2 x 8 JWPs

    I sized the processes based on the queue monitor. Nevertheless, in the recommendation of broadcom, there are much more DWPs. But the quoe of DWPs is always <5. Also the connections which are used by one DWP is usually not more then 7.




  • 4.  RE: Sizing of JWPs, DWPs, WPs + Session Timeouts

    Posted Jul 02, 2019 03:56 PM
    Hi

    Hmmm that seems ok to me (I would have bet that you have only 1 or 2 JWPs).
    To give it a try you could reduce DWPs by 2x4 and add 2x4 WPs  -- but honestly thats only a guess... :-(

    Do the logfiles indicate any connection issues?

    cheers, Wolfgang

    ------------------------------
    I know I do really know it!
    ------------------------------



  • 5.  RE: Sizing of JWPs, DWPs, WPs + Session Timeouts

    Posted Jul 03, 2019 12:50 PM
    There is a strange behaviour of JWP when starting up. Instead of using the assigned port, the jwp tries to take port by port starting from the lowest port. This causes several entries in the logs when starting a JWP


  • 6.  RE: Sizing of JWPs, DWPs, WPs + Session Timeouts

    Posted Jul 03, 2019 02:54 PM
    Thats normal (as far as I know)

    As you do not define a specific Port for a JWP it checks one by one if its occupied or not.

    cheers, Wolfgang

    ------------------------------
    I know I do really know it!
    ------------------------------



  • 7.  RE: Sizing of JWPs, DWPs, WPs + Session Timeouts
    Best Answer

    Posted Jul 04, 2019 08:39 AM

    Hello Michael,

    Are you reorganizing the MQ tables from within Automic or on DB Level?

    There are settings for MQ_BLOCK_COUNT and MQ_CHECK_TIME. You can disabe it and run the UC_REORG procedure from within DB.

    Please check your block count for all MQ tables. If you find them particularly large - they need to be reorganized as soon as possible. 

    This is not so related to the sizing but for the maintaining. If the tables blocks are too big you can experience performance degradation.

    MQ_BLOCK_COUNT

    The number of blocks after which the message queue is to be reorganized (Oracle databases only).

    Allowed values: 0 to 100000
    Default value: 64

    No
    MQ_CHECK_TIME

    The interval (seconds) at which message queues are checked (Oracle databases only).

    Allowed values: 0 to 1000000
    Default value: 600

    Value 0 means that no check is made.

    No


    ------------------------------
    Krum Ganev
    DXC Batch Scheduling SME
    ------------------------------