System data set failures
If an IMS system data
set fails because of an unrecoverable I/O error, you must scratch,
reallocate, and reformat the data set before you restart the system
using the /ERESTART command.
Message queue data set failures
Two types of problems can occur with message queue data sets: they can run out of space, or an I/O error can occur. In either case, you need to scratch and reallocate the problem data set, and increase its size if necessary.
Other system data set failures
You should make periodic backup copies of IMS system data sets so that if an error occurs, you can recover your data or recreate it from a backed up copy.
RECON data set recovery
If an I/O error occurs on a RECON data set, DBRC tries to use an available spare data set. DBRC then copies the healthy RECON data set to the spare data set, and then activates the spare data set.
Restoring RECON data sets if both are unusable
It is unlikely that both RECON data sets would be unusable. However, if this situation occurs, you must back up, delete, and redefine your RECON data sets, make an image copy of all applicable DBDSs, and close any open, out-of-date OLDSs.