Multi-threaded restore does not respect allocation mode
When restoring a Smalltalk full backup into a repository with a different
number of extents than the repository in which the original backup was
taken, the DBF_ALLOCATION_MODE should be used to determine how to distribute
the data. However, with balanced allocation mode, the multi-threaded restore
still put objects into the extents corresponding to the extents in which
they were in the original repository, or disproportionally favoring the
first extent, resulting in unbalanced final extent sizes.
Workaround
Using DBF_PRE_GROW results in balanced extent sizes, but the objects will
remain unbalanced. To produce well balanced results, use DBF_PRE_GROW =
TRUE and extent sizes specified to be slightly larger than the final extent
sizes, and after pregrowing, shut down and restart without extent sizes.
Last updated: 11/21/12