Hi David,
That is a great suggestion, except my partitions are foreign tables.
I wonder if I could figure out a single empty partition that’s just a normal
table... but I’m using hash partitioning, so I think I would lose a modulus.
Best,
Seamus
--
Seamus Abshere, SCEA
+1 (608) 772-0696
https
ed tables?
Best,
Seamus
PS. The awesome thing about this is that I'm table scanning a 270 million row,
600 column table in 1.2 seconds, which I never thought was possible.
PPS. I have no idea why 96 worked better than 64 (the number of cores I have),
but it did - 700% vs 400%.
On Sat, Feb 13
g for transaction cost). So
why doesn't it ask for more? Note that I've set max_parallel_workers=512, etc.
(postgresql.conf in my first message).
Here are full plans
https://gist.github.com/seamusabshere/1c1d869ee24145689f45e4124d8730d3
Best,
Seamus
On Fri, Feb 12, 2021, at 4:50 PM, Seamu
x_parallel_workers_per_gather = 512
max_wal_size = 4GB
max_worker_processes = 512
min_wal_size = 1GB
random_page_cost = 1.1
shared_buffers = 127232MB
shared_preload_libraries = 'cstore_fdw'
synchronous_commit=off
wal_buffers = 16MB
work_mem = 1628560kB
Best,
Seamus
--
Seamus Abshere, SCEA
https:
org/message-id/flat/CAMkU%3D1yV7WQLetrCVPqn%3DdTPdNzW3JD29ZsK0zJmgzO2tdcx-Q%40mail.gmail.com#aa3b3316cbe44fb3fd913ce019a86bb8
[2]
https://github.com/postgres/postgres/blame/master/src/backend/optimizer/path/costsize.c#L1056
--
Seamus Abshere, SCEA
+598 9954 9954
https://www.faraday.io
https:/
hi,
I created an index with text_pattern_ops because I want fast prefix search [1]
[2].
CREATE INDEX ON mytable USING BTREE (state, city, address text_pattern_ops);
(Note it's composite - I'm looking for prefixes on the address part only.) I
can see that it "works" (purely uses the index) for
l from subscribed where email not in (select tracks.ema...
(the "tracks" table doesn't have an "email" column, so the first query is just
meaningless)
Should there be a warning for the first query that you reference "tracks" in a
subquery but don't use any c
