BMP failures

If a BMP fails, DBCTL backs out any changes made by that BMP following the latest successful syncpoint. You must restart BMPs, because DBCTL does not restart them automatically.

The JCL used to restart BMPs depends on whether the checkpoint for the BMP is still on an OLDS available to DBCTL. If the BMP’s last checkpoint records are not in the OLDS, they will be in the SLDS, and you must add an IMSLOGR DD statement for the SLDS(s) containing the log records required to the BMP JCL. Guidance on the JCL needed to do this is in Database utilities in IMS product documentation.

There is an option to defer changes made to databases by backout of BMPs at emergency restart. If you specify NOBMP on the /ERESTART command, changes made to databases by BMPs are not backed out and all PSBs affected are stopped. Databases that were being updated by BMPs when the failure occurred are also stopped. You must then do batch backout for the databases that are stopped. (Batch backout will also backout the databases that were affected.) Be aware that using NOBMP may mean that the online DBCTL is restarted sooner, but it also delays data availability for the databases that were stopped by the BMP failure.