Improve last token native price sql query

[squadgazzz]

Currently, the query performs awfully for unpopular tokens. Also, the metrics were not updated properly. This PR fixes both issues. I was unable to reproduce it on staging, so another hotfix is required here to improve the tenderly alert execution time.

Execution plan before(auction_id is used first, which causes the performance issue):

+------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|QUERY PLAN                                                                                                                                                              |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|Limit  (cost=0.56..63.18 rows=1 width=62) (actual time=21566.534..21566.545 rows=0 loops=1)                                                                             |
|  ->  Index Scan Backward using auction_prices_pkey on auction_prices  (cost=0.56..2323157.04 rows=37102 width=62) (actual time=21566.533..21566.539 rows=0 loops=1)|
|        Index Cond: (token = '\xd684b049314ea296b9603f6ee3440380b08c8aa2'::bytea)                                                                                       |
|Planning Time: 0.417 ms                                                                                                                                                 |
|Execution Time: 21566.620 ms                                                                                                                                            |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
                                                                                                                                                                         

And after:

+--------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|QUERY PLAN                                                                                                                                                          |
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|Limit  (cost=0.56..4.50 rows=1 width=62) (actual time=0.186..0.187 rows=0 loops=1)                                                                                  |
|  ->  Index Scan using auction_prices_token_auction_id_idx on auction_prices  (cost=0.56..144221.68 rows=36690 width=62) (actual time=0.185..0.185 rows=0 loops=1)|
|        Index Cond: (token = '\xd684b049314ea296b9603f6ee3440380b08c8aa2'::bytea)                                                                                   |
|Planning Time: 1.618 ms                                                                                                                                             |
|Execution Time: 0.247 ms                                                                                                                                            |
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------+

[github-actions]

Reminder: Please update the DB Readme and comment whether migrations are reversible (include rollback scripts if applicable).

---

Caused by:

• database/sql/V079__auction_prices_token_index.sql

[sunce86]

Leaving a reminder to myself that here is another performance reason why we will probably have to keep auction_prices table, and erase auction prices from competition_auctions. (there are a few more places in the code)

[m-lord-renkse]

Really nice!!