Inbox (unsorted notes)

Interpreting buffers in analyze

Shared hit counters don’t necessarily correspond to sizes that would be read on a cold read (see this article).

The example given: an index scan within a nested loop will “hit” the block N (loop) times. I’m also wondering if this can happen with particularly scattered index scans (e.g. on updated_at where the entries are scattered all over the heap?)

Also noted from the Postgres docs (emphasis mine):

The buffer numbers displayed show the count of the non-distinct buffers hit, read, dirtied, and written for the given node and all of its child nodes.