NDB Concepts › Fix a Corrupted NDB

Fix a Corrupted NDB

An NDB full or region canceled event can corrupt an NDB. To recover the data, use the NDB ALTER OPT=BLDX command to rebuild the indexes.

Note: This procedure applies to an IPLOG file. Similar steps apply to all NDBs.

To fix a corrupted NDB

1. Create a work file. The size must relate to the number of records in the NDB.

   The following shows some sample JCL:

   //DEFWRK   EXEC  PGM=IDCAMS

   //SYSPRINT DD  SYSOUT=*
   //SYSIN    DD  *
    DELETE (hlq.TEST.NDBWORK) CL
    DEFINE CLUSTER (NAME(hlq.TEST.NDBWORK)      -  
                    MGMTCLAS(DEFAULT)           -  
                    STORCLAS(NMDPOOL)           -  
                    INDEXED                     -  
                    RECORDS(1000000 100000)     -  
                    SPEED                       -  
                    SHAREOPTIONS(2 3)           -  
                   )                            -  
           DATA    (NAME(hlq.TEST.NDBWORK.D)    -  
                    CONTROLINTERVALSIZE(4096)   -  
                    RECORDSIZE(1017 4089)       -  
                    FREESPACE(0 0)              -  
                    KEYS(4 0)                   -  
                   )                            -  
           INDEX   (NAME(hlq.TEST.NDBWORK.I)    -  
                    CONTROLINTERVALSIZE(1024))     
   /*
   

2. Issue the following command from OCS to allocate the work file to the region:

   ALLOCATE DD=NDBWORK DSN=hlq.TEST.NDBWORK DISP=SHR
   

   Issue the following command to open the work file as a VSAM file.

   UDBCTL OPEN=NDBWORK STRNO=3 BUFNI=5 BUFND=6
   

3. Issue the following command to remove the NDB bad-locked status:

   NDB PURGE IPLOG
   

4. Issue the following command to stop the NDB to prevent access:

   NBD STOP IPLOG IMM
   

5. Issue the NDB ALTER command to verify the indexes, as follows:

   NDB ALTER IPLOG OPT=CHKX DB WORK=NDBWORK DETAIL=YES SORT=100000
   

   Minimizing I/O on the work file by specifying the maximum value for sort memory reduces execution time.

6. Issue the following command to close and free the work file:

   UDBCTL CLOSE=NDBWORK
   UNALLOC DD=NDBWORK
   

7. Review the CHKX. If it shows an error, rebuild the indexes by redefining the work file to empty it. Then, use the BLDX option of NDB ALTER, by issuing the following command:

   NDB ALTER IPLOG OPT=BLDX DB WORK=NDBWORK DETAIL=YES SORT=100000
   

   Minimizing I/O on the work file by specifying the maximum value for sort memory reduces execution time.

   If the NDB ALTER BLDX action fails, the work file must be unallocated and the file must be redefined before you reissue the command.

8. Reallocate the NDB by actioning the IPFILES - TCP/IP File Specifications Customizer Parameter Group.