Hi,
TImescale makes use of inheritance in its partitioning implementation,
so we can't make use of the publish_via_partition_root publication
option during logical replication. We don't guarantee that the exact
same partitions exist on both sides, so that's a major roadblock for
our implementing logical subscription support, and by the same token
it's not possible to replicate out to a "standard" table.
If we were to work on a corresponding publish_via_inheritance_root
option, is there a chance that it'd be accepted, or is there some
other technical reason preventing it? In addition to Timescale, it
seems like other installations using extensions like pg_partman could
potentially make use of this, during online migrations from the old
style of partitioning to the new.
Some inheritance hierarchies won't be "partitioned" hierarchies, of
course, but the user can simply not set that replication option for
those publications. (Alternatively, I can imagine a system where an
extension explicitly marks a table as having a different "publication
root", and then handling that marker with the existing replication
option. But that may be overengineering things.)
WDYT?
--Jacob
