On Wed, Apr 2, 2025 at 10:53 PM Sami Imseih <samims...@gmail.com> wrote:
> > That seems like a very good location for this advice. But the current > > set of bullet points are all directed towards "... a general procedure > > for determining which indexes to create". I propose that a new paragrph, > > not a bullet point, be added towards the end of that section which > > addresses the options of steps to take before dropping an index. > > Something like the following: > > > Thoughts? > > This new feature provides the ability to experiment with indexes to > create ( or drop ), > so I don't think it deviates from the purpose of this paragraphs. > > This patch will provide the ability for the user to create an index as > initially > invisible and a GUC, use_invisible_index if set to TRUE to experiment with > the new index to see if it improves performance. So, I think providing this > pattern to experiment with a new index will fit nicely as a new > bulletpoint. > > > Thank you for the feedback and pointers Sami and Gurjeet. Good call on [0] being a good place for operational advice. I have gone ahead and removed the advice about "pg_stat_user_indexes.idx_scan" from doc/src/sgml/ref/alter_index.sgml and updated doc/src/sgml/indices.sgml to include a new bullet point with also a reference to use_invisible_index. Let me know how it sounds and if there is any feedback. Also, rebased. Thank you Shayon [0] https://www.postgresql.org/docs/current/indexes-examine.html
v15-0001-Introduce-the-ability-to-set-index-visibility-us.patch
Description: Binary data
