perf: update proto_checkpoint_view to use index

[NicholasBlaskey]

Description

proto_checkpoint_view was casting checkpoints_v2.uuid to a text meaning we couldn't use the index on checkpoints_v2.uuid so model queries were full scanning checkpoints.

Test Plan

TestCheckpointReturned should have this covered well

Commentary (optional)

Checklist

• Changes have been manually QA'd
• User-facing API changes need the "User-facing API Change" label.
• Release notes should be added as a separate file under docs/release-notes/.
  See Release Note for details.
• Licenses should be included for new code which was copied and/or modified from any external code.

Ticket

[netlify]

✅ Deploy Preview for determined-ui canceled.

Name	Link
🔨 Latest commit	b1fe644
🔍 Latest deploy log	https://app.netlify.com/sites/determined-ui/deploys/65bd432e01defb00080f4f93

[stoksc]

@NicholasBlaskey can you post the new plan?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (abd590d) 47.72% compared to head (b1fe644) 53.93%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8793      +/-   ##
==========================================
+ Coverage   47.72%   53.93%   +6.20%     
==========================================
  Files        1049      616     -433     
  Lines      167258    70734   -96524     
  Branches     2243        0    -2243     
==========================================
- Hits        79816    38147   -41669     
+ Misses      87284    32587   -54697     
+ Partials      158        0     -158     

Flag	Coverage Δ
backend	43.24% <ø> (+<0.01%)	⬆️
harness	63.19% <ø> (-1.13%)	⬇️
web	?

Flags with carried forward coverage won't be shown. Click here to find out more.

see 454 files with indirect coverage changes

[NicholasBlaskey]

new query

                                                                                       QUERY PLAN                                                                                       
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Nested Loop  (cost=13.11..64.55 rows=1 width=266) (actual time=2.409..2.434 rows=1 loops=1)
   ->  Nested Loop Left Join  (cost=2.73..54.09 rows=1 width=2461) (actual time=1.778..1.784 rows=1 loops=1)
         ->  Nested Loop Left Join  (cost=2.15..45.76 rows=1 width=1947) (actual time=1.741..1.745 rows=1 loops=1)
               ->  Nested Loop Left Join  (cost=1.71..37.74 rows=1 width=1730) (actual time=1.646..1.649 rows=1 loops=1)
                     ->  Nested Loop Left Join  (cost=1.42..29.91 rows=1 width=1555) (actual time=1.206..1.209 rows=1 loops=1)
                           ->  Nested Loop Left Join  (cost=1.00..21.94 rows=1 width=839) (actual time=1.191..1.194 rows=1 loops=1)
                                 ->  Nested Loop  (cost=0.57..17.73 rows=1 width=835) (actual time=1.160..1.162 rows=1 loops=1)
                                       ->  Nested Loop Left Join  (cost=0.14..9.28 rows=1 width=218) (actual time=0.014..0.016 rows=1 loops=1)
                                             ->  Seq Scan on model_versions  (cost=0.00..1.05 rows=1 width=204) (actual time=0.006..0.007 rows=1 loops=1)
                                                   Filter: (model_id = 1)
                                                   Rows Removed by Filter: 3
                                             ->  Index Scan using users_pkey on users  (cost=0.14..8.16 rows=1 width=18) (actual time=0.006..0.006 rows=1 loops=1)
                                                   Index Cond: (id = model_versions.user_id)
                                       ->  Index Scan using checkpoints_v2_uuid_key on checkpoints_v2 c  (cost=0.43..8.45 rows=1 width=633) (actual time=1.144..1.144 rows=1 loops=1)
                                             Index Cond: (uuid = model_versions.checkpoint_uuid)
                                 ->  Index Only Scan using idx_checkpoint_id_run_id on run_checkpoints rc  (cost=0.43..4.21 rows=1 width=20) (actual time=0.014..0.015 rows=1 loops=1)
                                       Index Cond: (checkpoint_id = c.uuid)
                                       Heap Fetches: 0
                           ->  Index Scan using trials_pkey on runs r  (cost=0.42..7.97 rows=1 width=720) (actual time=0.014..0.014 rows=1 loops=1)
                                 Index Cond: (id = rc.run_id)
                     ->  Index Scan using experiments_pkey on experiments e  (cost=0.29..7.83 rows=1 width=179) (actual time=0.439..0.439 rows=1 loops=1)
                           Index Cond: (id = r.experiment_id)
               ->  Index Scan using validations_trial_id_total_batches_run_id_unique on raw_validations v  (cost=0.44..8.01 rows=1 width=225) (actual time=0.012..0.013 rows=1 loops=1)
                     Index Cond: ((trial_id = r.id) AND (total_batches = ((c.metadata ->> 'steps_completed'::text))::integer))
                     Filter: (NOT archived)
         ->  Index Scan using steps_trial_id_total_batches_run_id_unique on raw_steps s  (cost=0.57..8.31 rows=1 width=522) (actual time=0.010..0.011 rows=1 loops=1)
               Index Cond: ((trial_id = r.id) AND (total_batches = ((c.metadata ->> 'steps_completed'::text))::integer))
               Filter: (NOT archived)
   ->  Subquery Scan on m  (cost=10.38..10.41 rows=1 width=191) (actual time=0.057..0.059 rows=1 loops=1)
         ->  GroupAggregate  (cost=10.38..10.40 rows=1 width=211) (actual time=0.048..0.049 rows=1 loops=1)
               Group Key: m_1.id, u.id
               ->  Sort  (cost=10.38..10.38 rows=1 width=207) (actual time=0.043..0.044 rows=1 loops=1)
                     Sort Key: u.id
                     Sort Method: quicksort  Memory: 25kB
                     ->  Nested Loop Left Join  (cost=0.14..10.37 rows=1 width=207) (actual time=0.011..0.014 rows=1 loops=1)
                           Join Filter: (mv.model_id = m_1.id)
                           ->  Nested Loop  (cost=0.14..9.31 rows=1 width=203) (actual time=0.008..0.010 rows=1 loops=1)
                                 ->  Seq Scan on models m_1  (cost=0.00..1.07 rows=1 width=185) (actual time=0.004..0.005 rows=1 loops=1)
                                       Filter: (id = 1)
                                       Rows Removed by Filter: 5
                                 ->  Index Scan using users_pkey on users u  (cost=0.14..8.16 rows=1 width=18) (actual time=0.003..0.003 rows=1 loops=1)
                                       Index Cond: (id = m_1.user_id)
                           ->  Seq Scan on model_versions mv  (cost=0.00..1.05 rows=1 width=8) (actual time=0.002..0.003 rows=1 loops=1)
                                 Filter: (model_id = 1)
                                 Rows Removed by Filter: 3
 Planning Time: 4.010 ms
 Execution Time: 2.714 ms
(47 rows)

[NicholasBlaskey]

old query

                                                                                        QUERY PLAN                                                                                                
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Gather  (cost=7427.00..508363.94 rows=16359 width=266) (actual time=440.571..1443.003 rows=1 loops=1)
   Workers Planned: 2
   Workers Launched: 2
   ->  Nested Loop Left Join  (cost=6427.00..505728.04 rows=6816 width=266) (actual time=1098.314..1431.299 rows=0 loops=3)
         ->  Nested Loop Left Join  (cost=6426.43..448657.93 rows=6816 width=2138) (actual time=1098.043..1431.017 rows=0 loops=3)
               ->  Parallel Hash Left Join  (cost=6425.99..393969.19 rows=6816 width=1921) (actual time=1098.014..1430.987 rows=0 loops=3)
                     Hash Cond: (r.experiment_id = e.id)
                     ->  Nested Loop Left Join  (cost=20.57..387545.87 rows=6816 width=1746) (actual time=1078.823..1411.793 rows=0 loops=3)
                           ->  Nested Loop Left Join  (cost=20.15..333237.67 rows=6816 width=1030) (actual time=1078.816..1411.786 rows=0 loops=3)
                                 ->  Hash Join  (cost=19.72..304518.72 rows=6816 width=1026) (actual time=1078.798..1411.766 rows=0 loops=3)
                                       Hash Cond: ((c.uuid)::text = (model_versions.checkpoint_uuid)::text)
                                       ->  Parallel Seq Scan on checkpoints_v2 c  (cost=0.00..292502.45 rows=1363245 width=633) (actual time=0.005..492.461 rows=1090596 loops=3)
                                       ->  Hash  (cost=19.70..19.70 rows=1 width=409) (actual time=0.125..0.132 rows=1 loops=3)
                                             Buckets: 1024  Batches: 1  Memory Usage: 9kB
                                             ->  Nested Loop  (cost=10.52..19.70 rows=1 width=409) (actual time=0.112..0.120 rows=1 loops=3)
                                                   ->  Subquery Scan on m  (cost=10.38..10.41 rows=1 width=191) (actual time=0.098..0.103 rows=1 loops=3)
                                                         ->  GroupAggregate  (cost=10.38..10.40 rows=1 width=211) (actual time=0.068..0.071 rows=1 loops=3)
                                                               Group Key: m_1.id, u.id
                                                               ->  Sort  (cost=10.38..10.38 rows=1 width=207) (actual time=0.057..0.060 rows=1 loops=3)
                                                                     Sort Key: u.id
                                                                     Sort Method: quicksort  Memory: 25kB
                                                                     Worker 0:  Sort Method: quicksort  Memory: 25kB
                                                                     Worker 1:  Sort Method: quicksort  Memory: 25kB
                                                                     ->  Nested Loop Left Join  (cost=0.14..10.37 rows=1 width=207) (actual time=0.032..0.037 rows=1 loops=3)
                                                                           Join Filter: (mv.model_id = m_1.id)
                                                                           ->  Nested Loop  (cost=0.14..9.31 rows=1 width=203) (actual time=0.023..0.025 rows=1 loops=3)
                                                                                 ->  Seq Scan on models m_1  (cost=0.00..1.07 rows=1 width=185) (actual time=0.013..0.014 rows=1 loops=3)
                                                                                       Filter: (id = 1)
                                                                                       Rows Removed by Filter: 5
                                                                                 ->  Index Scan using users_pkey on users u  (cost=0.14..8.16 rows=1 width=18) (actual time=0.007..0.007 rows=1 loops=3)
                                                                                       Index Cond: (id = m_1.user_id)
                                                                           ->  Seq Scan on model_versions mv  (cost=0.00..1.05 rows=1 width=8) (actual time=0.006..0.008 rows=1 loops=3)
                                                                                 Filter: (model_id = 1)
                                                                                 Rows Removed by Filter: 3
                                                   ->  Nested Loop Left Join  (cost=0.14..9.28 rows=1 width=218) (actual time=0.012..0.014 rows=1 loops=3)
                                                         ->  Seq Scan on model_versions  (cost=0.00..1.05 rows=1 width=204) (actual time=0.003..0.004 rows=1 loops=3)
                                                               Filter: (model_id = 1)
                                                               Rows Removed by Filter: 3
                                                         ->  Index Scan using users_pkey on users  (cost=0.14..8.16 rows=1 width=18) (actual time=0.007..0.007 rows=1 loops=3)
                                                               Index Cond: (id = model_versions.user_id)
                                 ->  Index Only Scan using idx_checkpoint_id_run_id on run_checkpoints rc  (cost=0.43..4.21 rows=1 width=20) (actual time=0.028..0.028 rows=1 loops=1)
                                       Index Cond: (checkpoint_id = c.uuid)
                                       Heap Fetches: 0
                           ->  Index Scan using trials_pkey on runs r  (cost=0.42..7.97 rows=1 width=720) (actual time=0.015..0.016 rows=1 loops=1)
                                 Index Cond: (id = rc.run_id)
                     ->  Parallel Hash  (cost=6183.52..6183.52 rows=17752 width=179) (actual time=18.723..18.724 rows=14202 loops=3)
                           Buckets: 65536  Batches: 1  Memory Usage: 9632kB
                           ->  Parallel Seq Scan on experiments e  (cost=0.00..6183.52 rows=17752 width=179) (actual time=0.004..6.354 rows=14202 loops=3)
               ->  Index Scan using validations_trial_id_total_batches_run_id_unique on raw_validations v  (cost=0.44..8.01 rows=1 width=225) (actual time=0.021..0.023 rows=1 loops=1)
                     Index Cond: ((trial_id = r.id) AND (total_batches = ((c.metadata ->> 'steps_completed'::text))::integer))
                     Filter: (NOT archived)
         ->  Index Scan using steps_trial_id_total_batches_run_id_unique on raw_steps s  (cost=0.57..8.31 rows=1 width=522) (actual time=0.014..0.017 rows=1 loops=1)
               Index Cond: ((trial_id = r.id) AND (total_batches = ((c.metadata ->> 'steps_completed'::text))::integer))
               Filter: (NOT archived)
 Planning Time: 2.267 ms
 Execution Time: 1443.133 ms
(56 rows)

[stoksc]

lgtm, checked it out locally and hit some apis to make sure there wasn't some random thing using that view that was untested.

[stoksc]

lgtm, checked it out locally and hit some apis to make sure there wasn't some random thing using that view that was untested.