With our backup techniques, it is a bit painful to deal with WAL files
when the number goes lower and starts re-using numbers already
archived, as when we upgrade to a new release using pg_dumpall piped
to psql. We can work around it, but it would keep our environment and
scripts simpler (and therefore less fragile) if we could count on a
later WAL file always having a name which sorts after any earlier
file.
I poked around the documentation, looking for a way that we could
force the new cluster to use a WAL file name higher than the last,
either within the some timeline or by bumping the timeline ID. The
only thing I found is pg_resetxlog, which seems like kind of a big
hammer to use on an otherwise sound database.
Did I miss anything in the documentation? Any suggestions?
-Kevin