Reduce regions_hard_delete contention #7960
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Recently, regions_hard_delete changed to use a CTE to update the size_used column of the crucible_dataset table. Unfortunately this had the side effect of causing a massive amount of contention: the CTE would 1) delete rows from the regions table 2) read from the regions table during the CTE 3) update the size_used column for _all_ crucible_dataset rows This would almost certainly cause each invocation of the CTE to contend with each other, as seen when doing disk deletes in parallel. This commit changes the CTE to: 1) delete rows from the regions table, returning the affected datasets 2) read from the regions table during the CTE 3) update the size_used column for affected crucible_dataset rows only. This was tested by using terraform to create and tear down 90 disks, with a parallelism setting of 10, 20, and 30. Before this change, this would not work as Nexus would inevitably return 500s. Fixes oxidecomputer#7952
smklein
reviewed
Apr 10, 2025
| .execute_async(&conn) | ||
| .await?; | ||
| let query = | ||
| regions_hard_delete::dataset_update_query(dataset_ids); |
Collaborator
There was a problem hiding this comment.
Yup, makes a lot of sense to have this be much more scoped!
| use uuid::Uuid; | ||
|
|
||
| /// Update the affected Crucible dataset rows after hard-deleting regions | ||
| pub fn dataset_update_query( |
Collaborator
There was a problem hiding this comment.
This could probably do with an EXPECTORATE + EXPLAIN test, to make it easier for future changes.
Comment on lines
42
to
49
| for (idx, dataset_id) in dataset_ids.into_iter().enumerate() { | ||
| if idx != 0 { | ||
| builder.sql(","); | ||
| } | ||
| builder.param().bind::<sql_types::Uuid, _>(dataset_id); | ||
| } | ||
|
|
||
| builder.sql( |
Collaborator
There was a problem hiding this comment.
Would it be possible to use:
builder.param().bind::<diesel::pg::sql_types::Array<sql_types::Uuid>, _>(dataset_ids)Instead of creating a new bind parameter for each individual index?
Contributor
Author
There was a problem hiding this comment.
I started with this, but hit a runtime error, something like couldn't map from uuid[] -> uuid - maybe I was doing it wrong? idk
Contributor
Author
There was a problem hiding this comment.
thread 'db::queries::regions_hard_delete::test::explainable' panicked at nexus/db-queries/src/db/queries/regions_hard_delete.rs:104:14:
Failed to explain query - is it valid SQL?: DatabaseError(Unknown, "invalid cast: uuid[] -> uuid")
Collaborator
There was a problem hiding this comment.
The following works for me:
- crucible_dataset.id IN (",
- );
-
- for (idx, dataset_id) in dataset_ids.into_iter().enumerate() {
- if idx != 0 {
- builder.sql(",");
- }
- builder.param().bind::<sql_types::Uuid, _>(dataset_id);
- }
-
- builder.sql(
- ")
+ crucible_dataset.id = ANY (",
+ )
+ .param()
+ .bind::<diesel::pg::sql_types::Array<sql_types::Uuid>, _>(dataset_ids)
+ .sql(
+ ")
Contributor
Author
There was a problem hiding this comment.
Nice, yeah - I missed how IN is not the same as = ANY, changed in fc37358
smklein
approved these changes
Apr 11, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Recently, regions_hard_delete changed to use a CTE to update the size_used column of the crucible_dataset table. Unfortunately this had the side effect of causing a massive amount of contention: the CTE would
This would almost certainly cause each invocation of the CTE to contend with each other, as seen when doing disk deletes in parallel.
This commit changes the CTE to:
This was tested by using terraform to create and tear down 90 disks, with a parallelism setting of 10, 20, and 30. Before this change, this would not work as Nexus would inevitably return 500s.
Fixes #7952