On 9 November 2018 at 18:45, Amit Langote <Langote_Amit_f8@lab.ntt.co.jp> wrote:
> As long as queries involve tuple routing that touches multiple not yet
> seen partitions, someone doing conflicting operations directly on multiple
> partitions in a transaction will have to be ready to see deadlocks.
> Maybe, we can document that.
Perhaps it's good enough. A guess there's at least a workaround of
doing LOCK TABLE <root_partitioned_table> in the CREATE INDEX /
TRUNCATE session.
--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services