Hi, > ... and it is claimed that autovacuum will never be triggered in order > to set hint bits, assuming we never modify the table again.
Actually I waited a bit and got a better EXPLAIN: ``` Index Only Scan using humidity_idx on humidity (cost=0.42..181.36 rows=1970 width=4) (actual time=0.372..16.869 rows=2904.00 loops=1) Index Cond: ((ts >= '2022-01-01 00:00:00'::timestamp without time zone) AND (ts < '2022-05-02 00:00:00'::timestamp without time zone) AND (city = 'M oscow'::text)) Heap Fetches: 0 Index Searches: 1 Buffers: shared hit=44 Planning Time: 0.520 ms Execution Time: 17.980 ms (7 rows) ``` This happens when CHECKPOINT starts: ``` 2025-04-03 18:36:23.435 MSK [209952] LOG: checkpoint starting: time ``` Interestingly it takes unusually long for my toy database: ``` 2025-04-03 18:40:53.082 MSK [209952] LOG: checkpoint complete: wrote 3522 buffers (5.4%), wrote 1 SLRU buffers; 0 WAL file(s) added, 0 removed, 5 recycled; write=269.463 s, sync=0.029 s, total=269.647 s; sync files=32, longest=0.004 s, average=0.001 s; distance=68489 kB, estimate=68489 kB; lsn=0/F4F3870, redo lsn=0/F167DD0 ``` There is nothing in between these two lines. To my humble knowledge, CHECKOINT shouldn't set hint bits and should take that long. At this point I don't know what's going on. This is `master` branch, b82e7eddb023. -- Best regards, Aleksander Alekseev