On 12/6/22 09:03, Tom Lane wrote:
Ron <ronljohnsonjr@xxxxxxxxx> writes:
You've set effective_cache_size to 83%, when it should be at most 20%.
I think you're confusing effective_cache_size with the size of
shared buffers. It's meant to be an estimate of the total disk
cache space available, in both shared buffers and OS caches.
In any case, it's a fairly spongy parameter. It doesn't correspond
to any allocation actually made or requested in Postgres; it just
serves to tweak the planner's guesses about how expensive repeated
indexscans are likely to be. 83% of available RAM might be too
much, but it's probably not causing any real problems.
The docs say that ECS is an estimate of how much cache one query can use,
no? If more than one query is running on a server at a time, then wouldn't
a huge ECS confuse things?
--
Angular momentum makes the world go 'round.
