Results 1 to 7 of 7
  1. #1
    Join Date
    Aug 2016
    Posts
    20
    Provided Answers: 1

    Unanswered: multiple instance with STMM ON

    Hi

    IF there are two instance with instance_memory to Automatic,

    Will each instance consider it has about 85% memory from OS. As db2pd -dbptmem shows for each instance its taking about 85%. Please confirm.

    regds
    paul

  2. #2
    Join Date
    Jul 2016
    Location
    Moscow
    Posts
    108
    Provided Answers: 13
    Hi Paul,

    Each instance doesn't consider that it "has" any amount of OS memory. STMM running on each database takes into account OS free memory before making any decision on adjusting memory used.
    Regards,
    Mark.

  3. #3
    Join Date
    Aug 2016
    Posts
    20
    Provided Answers: 1
    thanks mark.

    for multiple instance env, it is recommended to have instance_memory to hard value and database_memory to automatic and STMM ON.

    Coz recently we faced a issue with multiple instance, server was crashing with swapping and it used to happen very quick and it was not captured in db2pd coz of reboot.

    System admin said db2sysc of one of instance causing this. Both instance_memory was set to automatic. So if db2 only consider free memory it shld not go to swap right ? . may be wrong assumption. Please suggest

    regds
    paul

  4. #4
    Join Date
    Apr 2012
    Posts
    1,034
    Provided Answers: 18
    You may be exposed to an APAR related to leaking of memory. It may be unrelated to STMM. Post relevant facts (operating system details, DB2-server edition and fixpack details, whether TSA/hadr is involved or other cluster manager, the backup/logarchive solution), and the ulimits of the instance-owners (if Unix).

  5. #5
    Join Date
    Jul 2016
    Location
    Moscow
    Posts
    108
    Provided Answers: 13
    Paul,

    It's still not perfect and you may find some APARs on STMM fixed in the latest fixpacks only like this, for example:
    IT07070: STMM MAXIMUM TARGETS MAY BE TEMPORARILY TOO HIGH, SOME SWAPPING MAY OCCUR ON LINUX SYSTEMS
    You can try to evaluate the DB2_MEM_TUNING_RANGE registry variable trying to protect against such an issue in future, but there is no any guarantee.

    I've seen a number of times some "running out" application which suddenly started to consume enormous application heap size. Since APPL_MEMORY was automatic it led to OS swapping or even hang as in your case.
    To catch a problematic situation you can configure to fire db2 FODC with db2fodc - DB2 first occurrence data collection command with -detect option.
    It should help IBM support to investigate the root cause of the problem.
    Last edited by mark.bb; 10-20-16 at 13:01.
    Regards,
    Mark.

  6. #6
    Join Date
    Aug 2016
    Posts
    20
    Provided Answers: 1
    thanks mark,

    db2 10.1 fp 5
    aix 7.1


    IT01986 MEMORY LEAK HAPPENS WHEN FEDERATED SET TO 'Y'
    http://www-01.ibm.com/support/docvie...id=swg1IT01986

    Question

    Should stmmlog and diag log will give information if memory is not enough in OS and db2 trying to use more ? .

    regds
    paul

  7. #7
    Join Date
    Jul 2016
    Location
    Moscow
    Posts
    108
    Provided Answers: 13
    Paul,

    You can run the following commands periodically to check for such warnings:

    db2diag -g "funcname=sqloMemLogPoolConditions"
    db2diag -g "funcname=stmmCalcAutoScaleFactor"
    Regards,
    Mark.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •