> > > I would like to add a --freeze parameter to vacuumdb for use by the
> > > binary upgrade utility, and for symmetry with the existing VACUUM
> > > options; patch attached.
> >
> > Exactly what do you think the upgrade utility is going to do with it?
> > Surely not a database-wide VACUUM FREEZE, if we are hoping that upgrade
> > is going to be fast.
> >
> > As far as I can see this is a solution looking for a problem.
>
> I didn't go into the use-case. The way pg_migrator works is to copy the
> _schema_ from the old database and load it into the new database. We
> then need to run vacuum freeze on the schema-only databases because we
> then move pg_clog from the old database to the new one; so, it is
> needed, and it will not take long to run.
My first impulse was the same as Tom's, thanks for the explanation.
To the filled database case:
Would it make sense to enhance --table to allow wildcards and remove the
"cannot vacuum a specific table in all databases" check ?
One more question I have though is:
How do you make sure noone (e.g. autovacuum analyze)
unfreezes tuples after the vacuum freeze ?
Andreas
