Re: Draft for basic NUMA observability

On 4/3/25 15:12, Jakub Wartak wrote:
> On Thu, Apr 3, 2025 at 1:52 PM Tomas Vondra <tomas@vondra.me> wrote:
> 
>> ...
>>
>> So unless someone can demonstrate a use case where this would matter,
>> I'd not worry about it too much.
> 
> OK, fine for me - just 3 cols for pg_buffercache_numa is fine for me,
> it's just that I don't have cycles left today and probably lack skills
> (i've never dealt with arrays so far) thus it would be slow to get it
> right... but I can pick up anything tomorrow morning.
> 

OK, I took a stab at reworking/simplifying this the way I proposed.
Here's v24 - needs more polishing, but hopefully enough to show what I
had in mind.

It does these changes:

1) Drops 0002 with the pg_buffercache refactoring, because the new view
is not "extending" the existing one.

2) Reworks pg_buffercache_num to return just three columns, bufferid,
page_num and node_id. page_num is a sequence starting from 0 for each
buffer.

3) It now builds an array of records, with one record per buffer/page.

4) I realized we don't really need to worry about buffers_per_page very
much, except for logging/debugging. There's always "at least one page"
per buffer, even if an incomplete one, so we can do this:

   os_page_count = NBuffers * Max(1, pages_per_buffer);

and then

  for (i = 0; i < NBuffers; i++)
  {
      for (j = 0; j < Max(1, pages_per_buffer); j++)
      {
          ..
      }
  }

and everything just works fine, I think.


Opinions? I personally find this much cleaner / easier to understand.


regards

-- 
Tomas Vondra