Log stream data set migration and recall considerations

Logger allows both log steam offload and staging data sets to be migrated. Logger will automatically recall data sets when they are needed. You can migrate data sets using SMS or another protocol.

Some applications may not function properly if long delays are encountered while logger waits for a data set to be recalled. The IXCMIAPU keyword OFFLOADRECALL(NO) specifies that logger is not to recall the current log stream offload data set when performing offload processing for a given log stream. If the current log stream offload data set is migrated, logger will move to a new offload data set by allocating a new one and then continue the offload by writing to the new data set. While this may avoid the delay of recalling the offload data set, it may cause a waste of space on the DASD if the data set is recalled. Therefore care should be taken when using OFFLOADRECALL(YES) to size the data sets accordingly.

With OFFLOADRECALL(YES), logger will not be able to offload until the current offload data set can be recalled. If HSM is not available and either:

The current offload data set for one or more of the logstreams residing in the structure is migrated while the structure is being rebuilt, or
Logger is going through recovery for one of the logstreams whose current offload data set is migrated.

The following messages might appear:

IXG114A OFFLOAD IS NOT PROGRESSING ON sysname LOGSTREAM: logstreamname, STRUCTURE: structurename CHECK FOR OFFLOAD INHIBITORS WITHIN sysname.
IXG115A CORRECT THE OFFLOAD CONDITION ON sysname FOR structurename OR REPLY TASK=END TO END THE STRUCTURE TASK.

Similarly, if an IXGBRWSE request is made for log stream data that resides in a migrated data set, logger will not be able to process the request until the data set is recalled. Thus, if HSM is down or the recall takes a long time, browse delays may result. The delays are acceptable as long as the application can tolerate them.

IBM® strongly suggest that you do not migrate staging data sets and avoiding migrating offload data sets whenever possible. If migration is necessary because of DASD space constraints, migrate the low priority offload data sets for which associated applications can handle delays in accessing the data sets. Use LS_STORCLAS and STG_STORCLAS to specify on which pool or volumes logger offload and staging data sets can be allocated. Use LS_MGMTCLAS and STG_MGMTCLAS to specify if the data sets are migratable or not migratable.

Consider associating the STORCLAS with the MGMTCLAS so that migratable data sets are all allocated in the same pool or on the same volumes.