Examples of NGT Recover operations
The recovery scenarios show the different NGT Recover execution phases. NGT Recover analyzes all requests in the SYSIN data set and then constructs a plan that consists of the execution phases needed to accomplish the work requested. The following table describes the major execution phases.
The LOG INPUT phase reads the DB2 log records required for recovery and sends them to a sort task. If you specified MAXLOGS > 1, NGT Recover starts multiple subtasks to handle log file allocation. The main task processes log records. The subtasks handle allocation.
The MERGE phase reads sorted log records and applies them to page images, from an image copy or, in the case of BACKOUT and LOGONLY, from the space itself. This phase recovers table spaces or indexes from logs, extracts keys and sends them to a sort, reads the space for BACKOUT or LOGONLY recovery, writes output image copy data sets, writes output change accumulation files, and performs many other tasks. Multiple MERGE phases can run in parallel (using the MAXLSORT option).
The UNLOAD phase extracts the index keys from a table space or partition that is not being recovered (and has no MERGE phase) and sends those keys to a sort task. A MERGE phase and an UNLOAD phase should not occur for the same table space or table space partition.
The BUILD phase constructs indexes by using sorted keys extracted in a MERGE or UNLOAD phase.
The SNAP phase uses XBM or SUF to perform a data set snap to restore an Instant Snapshot copy.
This section contains the following topics:
- Multiple table space recovery and index rebuilds
- Parallel index key sorts and multitasking index rebuilds
- Multiple log sorts, MERGE phases, and parallel index key sorts with multitasking index rebuilds
- Multiple table space and index recoveries
- BACKOUT recovery
- Instant Snapshot recovery