Joachim Wieland <joe@mcknight.de> writes:
> SET enable_bitmapscan = 0;
> EXPLAIN SELECT conname FROM pg_constraint WHERE conrelid = 'clstr_tst'::regclass;
> QUERY PLAN
> ---------------------------------------------------------------
> Seq Scan on pg_constraint (cost=0.00..27.15 rows=1 width=64)
> Filter: (conrelid = 54538::oid)
> (2 rows)
Hm, well, that's why it doesn't want to use a seqscan, but why is the
estimate so high? I get 7.35 on my boxes, vs 8.27 (which does agree
with yours) for the indexscans. Stranger and stranger.
Would you try inserting a "vacuum verbose pg_constraint" into the test
as well? Maybe that will tell something relevant.
regards, tom lane