IT30242: SQLIUNDOADDKEY MAY CRASH THE DATABASE IF ANOTHER TRANSACTION IS PERFORMING A SMO ON THE SAME PAGE

Db2 Version 11.1 Mod 4 Fix Pack 5 for Linux, UNIX, and Windows
Db2 Version 11.1 Mod 4 Fix Pack 6 for Linux, UNIX, and Windows
Db2 Version 11.1 Mod 4 Fix Pack 6 for Linux, UNIX, and Windows

APAR status

• Closed as program error.

Error description

• Database System crash causes the production database damaged.
  User can not make new connection on database.
  
  Multiple FODC observed before DATABASE CRASH
  
  Db2diag.log error messages:
  2019-08-13-13.16.13.506658+000 I28539404E630         LEVEL:
  Severe
  PID     : 4079                 TID : 139808458401536 PROC :
  db2sysc 0
  INSTANCE: DB2INST1             NODE : 000            DB   :
  MYDB
  APPHDL  : 0-16862              APPID:
  X.X.X.X.47106.190813131543
  UOWID   : 1                    ACTID: 111068
  AUTHID  : DB2INST1             HOSTNAME: MYHOST1
  EDUID   : 231370               EDUNAME: db2agent (MYDB) 0
  FUNCTION: DB2 UDB, index manager, sqliUndoAddKey, probe:519
  RETCODE : ZRC=0x8510000A=-2062548982=SQLP_LFUL
            "Lock list full - SQL0912 Reason code 1"
            DIA8310C Lock list was full.
  
  Root Cause:
  
  When undoing a change of a page which another transaction is
  currently performing a structure modification operation (SMO)
  such as a page split on, the undo process will see the SMO bit
  and try to grab the tree lock again, which might hit the lock
  manager boundary and crash the database.
  
  The issue happened during undo, typically during undo we do not
  acquire any additional resources, ie. we should have acquired
  whatever lock during statement execution, not during undo .. or
  at least reserved the LRB via IN mode.
  
  But for index undo a special case could happen if another
  transaction is doing a structure modification operation(SMO) and
  holding a tree lock, the undo process would wait for that SMO
  finished by acquiring the tree lock to see if has been released
  by the SMO, which might hit the lock manager boundary and crash
  the database.
  

Local fix

• Increase the size of locklist
  https://www.ibm.com/support/knowledgecenter/en/SSEPGG_11.1.0/com
  .ibm.db2.luw.admin.config.doc/doc/r0000267.html
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * ALL                                                          *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * See Error Description                                        *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Upgrade to Db2 11.1 Mod 4 Fixpack 5 or higher                *
  ****************************************************************
  

Problem conclusion

• First fixed in Db2 11.1 Mod 4 Fixpack 5
  

Temporary fix

• See local fix or increase lock size list
  

Comments

APAR Information

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

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

Fix information

• Fixed component name

  DB2 FOR LUW

• Fixed component ID

  DB2FORLUW

Applicable component levels

• RB10 PSN

     UP