ICF recovery considerations for DB2 group buffer pool duplexing

This topic describes recovery for DB2 group buffer pool duplexing in an internal coupling facility (ICF) environment. DB2 group buffer pools (GBPs) allow us to share data between members of a DB2 data sharing group, while duplexing creates another instance of a GBP structure in another coupling facility; in the event of structure damage or connectivity loss to a primary GBP structure, the system can switch to the secondary structure.

Currently, our DB2 data sharing group consists of 12 members. All of our group buffer pools are enabled for duplexing (DUPLEX(ENABLED)), and all have three CFs designated in their preference list. We have two coupling facilities using ICF CPs on the same CPCs as two of the DB2 members, and our third coupling facility is a stand-alone model. For information on how group buffer pools were initially implemented in our environment, please refer to "DB2 Group Buffer Pool Duplexing", of the December 1999 edition of our test report.

Note that the following coupling facility failure and recovery example assumes the failure occurs on an ICF in a 2 CF environment, the second CF being a stand-alone; messages shown may be issued repeatedly.


Messages issued when the ICF fails

If the system with the ICF housing the primary GBP structure fails, you'll receive messages similar to the following:

				IXL158I PATH 72 IS NOW NOT-OPERATIONAL TO CUID: FFF4
				        COUPLING FACILITY 009672.IBM.02.000000046781
				        PARTITION: 1  CPCID: 00

				IXC518I SYSTEM JB0 NOT USING
				         COUPLING FACILITY  009672.IBM.02.000000046781
				         PARTITION: 1    CPCID: 00
				         NAMED CF2
				         REASON: CONNECTIVITY LOST.
				         REASON FLAG: 13300002.

				DSNB228I @DBZ1 DSNB5STS GROUP BUFFER POOL GBP8
				         CANNOT BE ACCESSED FOR READ_STGSTATS
				         MVS IXLCACHE REASON CODE=X'0C1C0C06'


				IXC522I REBUILD FOR STRUCTURE DSNDB1G_GBP0 IS BEING STOPPED
				        TO SWITCH TO THE NEW STRUCTURE DUE TO
				        LOSS OF CONNECTIVITY TO THE STRUCTURE


				DSNB303E @DBG1 DSNB1REE A LOSS OF CONNECTIVITY WAS
				         DETECTED
				         TO GROUP BUFFER POOL GBP0


				DSNB744I @DBG1 DSNB1GBR DUPLEXING IS BEING STOPPED
				         FOR GROUP BUFFER POOL GBP0
				         SWITCHING TO SECONDARY
				         REASON = LOSSCONNPRI
				         DB2 REASON CODE = 00000000

				IXC101I SYSPLEX PARTITIONING IN PROGRESS FOR J90 REQUESTED BY
				        XCFAS.
				        REASON: SFM STARTED DUE TO STATUS UPDATE MISSING
				

Messages issued after the ICF system is partitioned from the sysplex

After the failing system that housed the ICF is partitioned out of the sysplex, the group buffer pools transition from duplex to simplex mode:

				IXC105I SYSPLEX PARTITIONING HAS COMPLETED FOR J90
				        - PRIMARY REASON: SYSTEM REMOVED
				        BY SYSPLEX FAILURE MANAGEMENT BECAUSE
				        ITS STATUS UPDATE WAS MISSING
				         - REASON FLAGS: 000100

				DSNB745I @DBG1 DSNB1GBR THE TRANSITION BACK TO
				          SIMPLEX MODE HAS COMPLETED FOR
				          GROUP BUFFER POOL GBP0


				IXL014I IXLCONN REBUILD REQUEST FOR STRUCTURE DSNDB1G_GBP0
				        WAS SUCCESSFUL.  JOBNAME: DBG1DBM1 ASID: 00FD
				        CONNECTOR NAME: DB2_DBG1 CFNAME: CF2

				IXC538I DUPLEXING REBUILD OF STRUCTURE DSNDB1G_GBP0
				        WAS NOT INITIATED BY MVS.
				        REASON: NO COUPLING FACILITY PROVIDED BETTER OR
				                EQUIVALENT CONNECTIVITY
				

Messages issued when the secondary GBP structure becomes available

Since we specify DUPLEX(ENABLED) for our group buffer pools in our CFRM policy, our group buffer pools are automatically rebuilt in duplex mode when the secondary coupling facility structure becomes available. You'll see messages similar to the following as this occurs:

				IXC517I SYSTEM Z1 ABLE TO USE
				         COUPLING FACILITY  009672.IBM.02.000000046781
				         PARTITION: 1    CPCID: 00
				          NAMED CF2

				IXC536I DUPLEXING REBUILD OF STRUCTURE DSNDB1G_GBP0
				        INITIATED.
				        REASON: GAINED CONNECTIVITY TO A COUPLING FACILITY

				DSNB740I @DBH1 DSNB1RBQ ATTEMPTING TO ESTABLISH
				         DUPLEXING FOR
				         GROUP BUFFER POOL GBP0
				         REASON = POLICY

				IXC529I DUPLEX REBUILD NEW STRUCTURE DSNDB1G_GBP0
				        IS BEING ALLOCATED IN COUPLING FACILITY CF2.
				        OLD STRUCTURE IS ALLOCATED IN COUPLING FACILITY CF1.
				        REBUILD START REASON: MVS INITIATED BASED ON POLICY


				IXL015I REBUILD NEW STRUCTURE ALLOCATION INFORMATION FOR
				        STRUCTURE DSNDB1G_GBP0, CONNECTOR NAME DB2_DBG1
				         CFNAME     ALLOCATION STATUS/FAILURE REASON
				         --------   ---------------------------------
				           CF1        RESTRICTED BY REBUILD OTHER
				           CF2        STRUCTURE ALLOCATED

				IXC521I REBUILD FOR STRUCTURE DSNDB1G_GBP0
				        HAS REACHED THE DUPLEXING ESTABLISHED PHASE

				DSNB742I @DBH1 DSNB1GBR DUPLEXING HAS BEEN
				         SUCCESSFULLY ESTABLISHED FOR
				         GROUP BUFFER POOL GBP0
				

Recovery completes

Now, because we specified DUPLEX(ENABLED) in our CFRM policy, our group buffer pools are automatically allocated once the failing ICF comes back online.

Note that table spaces do not require manual recovery when you have group buffer pools set up in duplex mode.