The impact of archiving logs in a data sharing group
In data sharing, the Db2 RECOVER utility needs log records from every member that has changed the object that needs to be recovered.
- Archive to disk without DFSMShsm to migrate the data sets from disk
to tape.
No major impact on performance is experienced for this type of archive. But you need enough disk space to hold archive logs, and the disk devices must be shared (accessible) by all members in a data sharing group. Because DFSMShsm or its equivalent is not used, you must manage disk space carefully to avoid running out of space.
- Archive to disk with DFSMShsm.
DFSMShsm can do automatic space and data availability management among storage devices in a system. DFSMShsm can migrate the archive on disk to less expensive storage (such as tape), and recall the archive back to disk when needed.
Using DFSMShsm, a particular RECOVER job needs only one tape unit to recall migrated archive data sets. If the archive data sets have been migrated, recovery time might be adversely affected, because the recalls of the migrated archive data sets are done one at a time from the member running the RECOVER job. For example, a RECOVER job started on DB1A might need log data sets from DB1A, DB2A, and DB3A. DB1A sends the recall requests to DFSMShsm one at a time for the tapes needed for recovery.
- Archive to tape.
The RECOVER job needs at least one tape unit for each member whose archived log records are to be merged. (More might be needed if you run more than one recover at the same time for different partitions of a partitioned table space.) Therefore, do not archive logs from more than one system to the same tape.
Recommendation: For data sharing, avoid using tape archive logs for data recovery.If you must archive to tape, ensure that the value for READ TAPE UNITS on installation panel DSNTIPA for each member is high enough to handle anticipated recovery work. For example, if you have eight members, each member should specify at least eight tape drives. You need more if you run more than one recovery job at the same time on a given member, or if multiple members run recovery jobs at the same time.
If there are not enough tape units to do the recovery, Db2 can possibly deadlock. If this happens, use the command SET ARCHIVE to increase the number of tape units that can be used.
Also, ensure that you specify 0 for the DEALLOC PERIOD parameter on installation DSNTIPA to avoid making an archive tape inaccessible to other members of the data sharing group. (If you intend to run all RECOVER jobs from a single member, this suggestion does not apply.)