On Wed, Jan 09, 2019 at 10:33:00AM +0900, Masahiko Sawada wrote:
> Since pg_(total)_relation_size() returns 0 for parent table the
> specifying the parent table to vacuumdb with --min-relation-size
> always does nothing. Maybe we will need to deal with this case when a
> function returning whole partitoned table size is introduced.
Good point. I am not sure if we want to go down to having a size
function dedicated to partitions especially as this would just now be
a wrapper around pg_partition_tree(), but the size argument with
partitioned tables is something to think about. If we cannot sort out
this part cleanly, perhaps we could just focus on the age-ing
parameters and the other ones first? It seems to me that what is
proposed on this thread has value, so we could shave things and keep
the essential, and focus on what we are sure about for simplicity.
--
Michael