PM56363: DISTRIBUTED SERVER USES TOO MUCH TCB CPU IN XXXXDIST ADDRESS SPACE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• DB2DDF defect pm56363 dpm56363
  CICS transactions routed through a DB2 for z/OS requester to
  a DB2 for z/OS server began slowing down.  The distributed
  address space was found to have been using too much TCB CPU.
  Error was traced to RACF calls at the server that results in
  ACEEs not being properly freed by DB2 for z/OS. The
  accumlation of ACEEs causes performance degradation during
  authentication processing (increased TCB time will be noted
  and associated to the xxxxDIST address space).
  ********************************************************
  Additional Symptoms and Keywords:
    PERF PERFORMANCE LP LOOP WAIT WT
    ACEE CCBACEE
    CICS HIGH CPU PERFORMANCE L8 TCB L8TCB
    CICS IMS NewUser New User signon sign-on sign on
  

Local fix

• The only way to clean up the accumulated ACEEs is to stop DB2,
  which terminates the xxxxDIST address space. DB2 can then be
  restarted.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All DB2 users.                               *
  ****************************************************************
  * PROBLEM DESCRIPTION: Performance degradation and excessive   *
  *                      CPU consumption occurs at a DB2 10      *
  *                      for z/OS subsystem when processing      *
  *                      authentication related events.          *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  Performance degradation and excessive CPU consumption occurs
  at a DB2 10 for z/OS subsystem when processing authentication
  related events.
  This problem mostly impacts Distributed Data Facility (DDF)
  related environments, especially where a DB2 10 for z/OS
  subsystem behaves as a server relative to remote DB2 for z/OS
  requester system CICS or IMS based applications.
  When remote application environments reuse their connections
  to a DB2 for z/OS server, DB2 may be required to drive SIGNON
  exit processing to reflect the new user of the connection. DB2
  for z/OS SIGNON exit processing will attempt to free the RACF
  ACEE that is associated with the prior user of the connection.
  DB2 10 for z/OS SIGNON exit processing logic was converted to
  64-bit addressing mode but did not consider that RACF will
  still only accept a 31-bit ACEE address. The result is that the
  ACEE is not really freed during SIGNON exit processing which
  causes an accumulation of ACEEs that will be known to RACF.
  Over time, the accumulation of ACEEs causes performance
  degradation of RACF authentication processing due to its effort
  to obtain a new ACEE. Users may also notice increased CPU TCB
  time, especially associated with the ssnmDIST address space.
  This is especially prevalent in cases where a password or
  PassTicket is provided, which must therefore be authenticated,
  which provides for more opportunity for an ACEE to be available
  during SIGNON exit processing.
  

Problem conclusion

• DB2 10 for z/OS SIGNON exit processing is changed to provide
  a 31-bit address to RACF when freeing the ACEE.
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

  UK76060

Modules/Macros

• DSN3AUSI
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RA10 PSY UK76060

     UP12/02/25 P F202

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.