chore(issues): Improve FileIOMainThreadDetector performance #82903
+53
−3
Conversation
This file contains 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
FileIOMainThreadDetector is our most time consuming detector: https://sentry.sentry.io/traces/?groupBy=span.description&interval=15m&mode=aggregate&project=1&query=span.op%3Afunction%20span.description%3Arun_detector_on_data.%2A&statsPeriod=24h&visualize=%7B%22chartType%22%3A1%2C%22yAxes%22%3A%5B%22sum%28span.duration%29%22%5D%7D This eliminates the need to recompile these glob patterns potentially for every transaction event and adds a synthetic benchmark to measure the (~4x) improvement: Before: ``` $ ./bin/benchmark_detectors 1,000,000 ops 7.005 s 142,758.90 ops/s ``` After: ``` $ ./bin/benchmark_detectors 1,000,000 ops 1.806 s 553,602.94 ops/s ```
github-actions
bot
added
the
Scope: Backend
mrduncan
commented
Jan 3, 2025
| def main(): | ||
| settings = get_detection_settings() | ||
|
|
||
| # 10 events: 1 ignored, 1 matching, and 8 ignored |
There was a problem hiding this comment.
These all came directly from tests/sentry/utils/performance_issues/test_file_io_on_main_thread_detector.py
mrduncan
commented
Jan 3, 2025
| @@ -117,7 +118,10 @@ class FileIOMainThreadDetector(BaseIOMainThreadDetector): | |||
|
|
|||
| __slots__ = ("stored_problems",) | |||
|
|
|||
| IGNORED_LIST = {"*.nib", "*.plist", "*kblayout_iphone.dat"} | |||
| IGNORED_LIST = { | |||
| re.compile(fnmatch.translate(pattern)) | |||
There was a problem hiding this comment.
sentry.utils.glob.glob_match is documented as "A beefed up version of fnmatch.fnmatch" but would appreciate any insights into ways they might diverge.
wedamija
approved these changes
Jan 6, 2025
There was a problem hiding this comment.
Nice fix! Since you have the benchmark, it could be worth seeing if compiling into a single regex would be faster, like
"|".join(fnmatch.translate(pattern) for pattern in patterns)
Alternatively, these are just suffix matches on files, so would it be even faster to just do
IGNORED_SUFFIX_LIST = [".nib", ".plist", "kblayout_iphone.dat"]
file_path.endswith(suffix) for suffix in IGNORED_SUFFIX_LIST)
```?
This is ~10-15% faster and more importantly it's a lot simpler to understand and reason about. Before (with pre-compiled multiple regexes) ``` $ ./bin/benchmark_detectors 1,000,000 ops 1.782 s 561,138.73 ops/s ``` After ``` $ ./bin/benchmark_detectors 1,000,000 ops 1.556 s 642,830.58 ops/s ```
|
Great idea - we're probably reaching the limits of the accuracy of this benchmark but I do like the idea of moving to a suffix check since it's also way simpler to reason about: Pre-compiled multiple regexes: Pre-compiled single regex: File path ends with: |
wedamija
approved these changes
Jan 6, 2025
❌ 3 Tests Failed:
View the top 3 failed tests by shortest run time
To view more test analytics, go to the Test Analytics Dashboard |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
FileIOMainThreadDetector is our most time consuming detector: https://sentry.sentry.io/traces/?groupBy=span.description&interval=15m&mode=aggregate&project=1&query=span.op%3Afunction%20span.description%3Arun_detector_on_data.%2A&statsPeriod=24h&visualize=%7B%22chartType%22%3A1%2C%22yAxes%22%3A%5B%22sum%28span.duration%29%22%2C%22count%28span.duration%29%22%2C%22p95%28span.duration%29%22%2C%22p99%28span.duration%29%22%5D%7D
This eliminates the need to recompile these glob patterns potentially for every transaction event and adds a synthetic benchmark to measure the (~4x) improvement:
Before:
After: