[MCC-1204770] Improve accessible objects pgsql function performance when filtering #201
Conversation
| class UpdateAccessibleObjectsFilterPerformanceGenerator < Rails::Generators::Base | ||
| source_root File.expand_path('../../../migrations', __FILE__) | ||
|
|
||
| def generate_update_accessible_objects_filter_performance_migration |
There was a problem hiding this comment.
Just out of curiosity, why do we do all this migration generation manually instead of the standard rails gem_name:install:migrations?
There was a problem hiding this comment.
Simple as this is the pattern the gem has been using since many years ago.
| 'INSERT INTO t_user_attribute_filtered_ids ' || | ||
| 'SELECT ua.child_id ' || | ||
| 'FROM t_user_attribute_ids ua ' || | ||
| 'JOIN policy_elements pe ' || | ||
| ' ON pe.id = ua.child_id ' || | ||
| 'WHERE %s', | ||
| filter_conditions |
There was a problem hiding this comment.
This seems a lot better than the deleting and NOT IN clause. 👍
| EXCEPTION | ||
| WHEN read_only_sql_transaction THEN | ||
| RETURN QUERY SELECT * FROM pm_accessible_objects_for_operations_cte(user_id, operations, field, filters); |
There was a problem hiding this comment.
This exception is raised when attempting to create a temporary table from a physical read replica. It's a weak fallback if the app is placed in read only mode.
There was a problem hiding this comment.
Oh, interesting. How often is pm_accessible_objects_for_operations_cte used? Does it exist solely for RR fallback?
|
notes for other reviewers: https://github.com/mdsol/kjerinsky-mdsol-notes/tree/notes/2024/mcc/MCC-1189079-investigate-slow-priv (link in desc is broken) |
|
That would be fine. We would just need to move the specs over or create new ones to test it. Bonus would be that I could add scientist experiments. |
|
besides that the query LGTM, my main question was answered in your notes: i lean toward just defining the func in dalton itself and not messing around w this gem for this but i dont feel too strongly and dont wanna create extra work for you |
|
per discussion earlier today, seems like we go with keeping it in the gem for now. merging. |
ejinotti-mdsol
deleted the
feature/accessible-objects-filter-performance
branch
MCC-1204770
Notes
Changes:
First statement selects by
DISTINCT. Thechild_idwas duplicated at times in significant numbers. A very marginal difference in ms at most if at all.Altered the filter statement to be more efficient. In the reproducible example, we reduced an 8+ minute call down to 180 ms.
@mdsol/platform-scalability