IZ65848: PAGE CLEANING PERFORMANCE ISSUE DUE TO TEMPORARY OBJECT DIRTY PAGES.

DB2 Version 9.1 Fix Pack 9  for Linux, UNIX and Windows
DB2 Version 9.1 Fix Pack 10  for Linux, UNIX and Windows
DB2 Version 9.1 Fix Pack 11  for Linux, UNIX and Windows
DB2 Version 9.1 Fix Pack 12  for Linux, UNIX and Windows

APAR status

• Closed as program error.

Error description

• Page cleaning often occurs in bufferpools to reduce the active
  log space being held up by dirty pages.  In doing so, the page
  cleaners might encounter some pages from temporary object that
  are currently in use.  On platforms without asynchronous IO, the
  page cleaners will wait until the temporary pages are released.
  During this period the page cleaners will stop all page cleaning
  activities, therefore allowing the dirty pages in the bufferpool
  to build up.  If the temporary pages are not released soon
  enough then this could cause the active log space to accumulate,
  even past the softmax settings.
  
  Since the temporary pages do not play a part in recovery,
  cleaning them out will not help with reducing the active log
  space, therefore it is not necessary to clean out the temporary
  pages.  The fix is to prevent the page cleaners from waiting on
  any temporary pages that are currently in use.
  
  Symptoms:
  1. The following message appears in the db2diag.log
  
  2009-11-24-10.37.49.841169+060 E14193576A697      LEVEL: Warning
  PID     : 15055                TID  : 1           PROC :
  db2agent (TEST)
  0
  INSTANCE: TEST               NODE : 000         DB   : TEST
  APPHDL  : 0-2114               APPID:
  10.21.22.58.51460.091124093715
  AUTHID  : TEST
  FUNCTION: DB2 UDB, data protection services, sqlpgResSpace,
  probe:1660
  MESSAGE : ADM1822W  The active log is being held by dirty pages.
  This is not an error, but database performance may be impacted.
  If possible, reduce the database work load.  If this problem
  persists,  either decrease the SOFTMAX and/or increase the
  NUM_IOCLEANERS DB configuration parameters.
  
  Also check if the active log space held by dirty pages
  substantially exceeds the softmax settings.
  
  2. SET WRITE SUSPEND takes a long time to finish.  The stack
  will show that it is waiting for pausing the page cleaners, and
  the page cleaners are waiting on page latches.
  

Local fix

• Workaround:
  1. Avoid using long running statements with temporary objects
  2. Use a large sortheap to avoid spilling
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All                                                          *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * Page cleaning often occurs in bufferpools to reduce          *
  * theactivelog space being held up by dirty pages.  In doing   *
  * so, thepagecleaners might encounter some pages from          *
  * temporary objectthatare currently in use.  On platforms      *
  * without asynchronous IO,thepage cleaners will wait until the *
  * temporary pages arereleased.During this period the page      *
  * cleaners will stop all pagecleaningactivities, therefore     *
  * allowing the dirty pages in thebufferpoolto build up.  If    *
  * the temporary pages are not released soonenough then this    *
  * could cause the active log space toaccumulate,even past the  *
  * softmax settings.Since the temporary pages do not play a     *
  * part in recovery,cleaning them out will not help with        *
  * reducing the active logspace, therefore it is not necessary  *
  * to clean out thetemporarypages.  The fix is to prevent the   *
  * page cleaners from waitingonany temporary pages that are     *
  * currently in use.Symptoms:1. The following message appears   *
  * in the db2diag.log2009-11-24-10.37.49.841169+060             *
  * E14193576A697      LEVEL:WarningPID    : 15055               *
  *  TID  : 1          PROC :db2agent (TEST)0INSTANCE: TEST      *
  *         NODE : 000        DB  : TESTAPPHDL  : 0-2114         *
  *      APPID:10.21.22.58.51460.091124093715AUTHID  :           *
  * TESTFUNCTION: DB2 UDB, data protection services,             *
  * sqlpgResSpace,probe:1660MESSAGE : ADM1822W  The active log   *
  * is being held by dirtypages.This is not an error, but        *
  * database performance may beimpacted.If possible, reduce the  *
  * database work load.  If this problempersists,  either        *
  * decrease the SOFTMAX and/or increase theNUM_IOCLEANERS DB    *
  * configuration parameters.Also check if the active log space  *
  * held by dirty pagessubstantially exceeds the softmax         *
  * settings.2. SET WRITE SUSPEND takes a long time to finish.   *
  * The stackwill show that it is waiting for pausing the page   *
  * cleaners,andthe page cleaners are waiting on page latches.   *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Upgrade to version db2 V9.1 FP9                              *
  ****************************************************************
  

Problem conclusion

• First fixed in db2 V9.1 FP9
  

Temporary fix

• n/a
  

Comments

• n/a
  

APAR Information

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

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

  IZ65866 IC64928

Fix information

• Fixed component name

  DB2 UDB ESE HP-

• Fixed component ID

  5765F4103

Applicable component levels

• R910 PSY

     UP