[Datastore] Fix ResourceCache table caching key bug #9085
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
gtopper
reviewed
Dec 18, 2025
Collaborator
gtopper
left a comment
gtopper
left a comment
There was a problem hiding this comment.
Looks great. Would be good to add a regression unit test.
gtopper
reviewed
Dec 21, 2025
tests/datastore/test_datastores.py
Outdated
Comment on lines
403
to
406
| # Act - First call creates the table | ||
| first_result = cache.get_table(test_uri) | ||
|
|
||
| # Assert - Table created and cached under original URI (not parsed path) |
Collaborator
There was a problem hiding this comment.
These comments are pretty unnecessarily verbose. We know the test will set things up, do something, and make an assertion...
alxtkr77
force-pushed
the
ML-11518-fix-table-cache-key
branch
from
e53b00b to
767f27b
Compare
gtopper
approved these changes
Dec 21, 2025
tests/datastore/test_datastores.py
Outdated
Comment on lines
389
to
391
| Regression test for ML-11518: parse_path() was overwriting the uri variable, | ||
| causing tables to be stored under the parsed path instead of the original URI. | ||
| This led to cache misses on subsequent calls with the same URI. |
Collaborator
There was a problem hiding this comment.
We probably don't need an in-depth recap here.
added 2 commits
December 21, 2025 12:38
Fix bug where parse_path() overwrote the uri variable, causing tables to be stored under the parsed path instead of the original URI. This caused cache misses on subsequent get_table() calls with the same URI, leading to duplicate Table objects being created. The fix preserves the original URI for the cache key while using the parsed path for the Table constructor, ensuring proper cache hits.
Test verifies that ResourceCache.get_table() caches tables under the original URI, not the parsed path. This prevents cache misses and duplicate Table objects on subsequent calls with the same URI.
alxtkr77
force-pushed
the
ML-11518-fix-table-cache-key
branch
from
767f27b to
808169d
Compare
Refactor test based on reviewer feedback: - Move imports to file top instead of inline in function - Remove verbose AAA comments (Arrange/Act/Assert) - Remove internal cache assertion (_tabels check) in favor of testing observable behavior (Table not recreated on second call) Reference: ML-11518
alxtkr77
force-pushed
the
ML-11518-fix-table-cache-key
branch
from
808169d to
4793685
Compare
gtopper
approved these changes
Dec 21, 2025
alxtkr77
added a commit
to alxtkr77/mlrun
that referenced
this pull request
Dec 21, 2025
## Summary - Fix bug in `ResourceCache.get_table()` where `parse_path(uri)` overwrote the `uri` variable - Tables were being stored under the parsed path instead of the original URI, causing cache misses - This led to duplicate Table objects being created on subsequent calls with the same URI ## Root Cause In `store_resources.py`, the code was: ```python endpoint, uri = parse_path(uri) # uri gets overwritten! self._tabels[uri] = Table(uri, ...) # stored under wrong key ``` The fix preserves the original URI for the cache key: ```python endpoint, path = parse_path(uri) # uri preserved self._tabels[uri] = Table(path, ...) # stored under correct key ``` ## Related Issue [ML-11518](https://jira.iguazeng.com/browse/ML-11518) ## Test Plan - Existing unit tests pass - Verified fix addresses table caching behavior during model monitoring drain cycles [ML-11518]: https://iguazio.atlassian.net/browse/ML-11518?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ --------- Co-authored-by: Alex Toker <[email protected]>
alxtkr77
added a commit
that referenced
this pull request
Dec 21, 2025
## Summary - Fix bug in `ResourceCache.get_table()` where `parse_path(uri)` overwrote the `uri` variable - Tables were being stored under the parsed path instead of the original URI, causing cache misses - This led to duplicate Table objects being created on subsequent calls with the same URI Backport of #9085 Co-authored-by: Alex Toker <[email protected]>
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.
Summary
ResourceCache.get_table()whereparse_path(uri)overwrote theurivariableRoot Cause
In
store_resources.py, the code was:The fix preserves the original URI for the cache key:
Related Issue
ML-11518
Test Plan