As the OP, I'll just note that my organization would definitely find use for a parallel migrator tool as long as it supported doing a selection of tables (i.e. -t / -T) in addition to the whole database and it supported or we were able to patch in an option to cluster as part of the migration (the equivalent of something like https://github.com/tgarnett/postgres/commit/cc320a71 ).
Tim
On Wed, Apr 24, 2013 at 5:47 PM, Joachim Wieland <joe@mcknight.de> wrote:
> What might make sense is something like pg_dump_restore which would have > no intermediate storage at all, just pump the data etc from one source > to another in parallel. But I pity the poor guy who has to write it :-)
hmm pretty sure that Joachims initial patch for parallel dump actually had a PoC for something very similiar to that...
That's right, I implemented that as an own output format and named it "migrator" I think, which wouldn't write each stream to a file as the directory output format does but that instead pumps it back into a restore client.
Actually I think the logic was even reversed, it was a parallel restore that got the data from internally calling pg_dump functionality instead of from reading files... The neat thing about this approach was that the order was optimized and correct, i.e. largest tables start first and dependencies get resolved in the right order.
I could revisit that patch for 9.4 if enough people are interested.