> You need to set $PGDATA before running the script. And psql,pg_ctl and
> pg_resetxlog need to be in $PATH. After running the script, restart
> postmaster and run "SELECT * FROM t2". There should be one row in the
> table, but it's empty.
I've tried this script on "postgres (PostgreSQL) 8.3devel", and found that
T2 is not empty after recovery(just as it should be)---but the latest
version
act just like what you said.
Then I see how cluster is done, and found that:
In "postgres (PostgreSQL) 8.3devel", unlike AlterTableSetTablespace (which
copys the whole relation block-by-block, and doesn't use wal under
non-archiving
mode), Cluster copys the relation row-by-row(simple_heap_insert), which
always uses wal regardless of archiving mode. As wal exists, recovery will
cope with everything rightly.
The latest version acts differently probably because that it removes wal of
cluser
under non-archiving mode.
So the conclusion is: we can replace wal mechanism with smgrimmedsync only
if
relfilenode is not allowed to be reused, but this's not true, so what we
should
keep wal.
Is it right?
