Reducing noise by improving Issue grouping #66319
Replies: 3 comments 7 replies
-
Slightly unrelated but one of the problems, I have with the "similar issues" view is that it is a little cluttered. "Issues with similar stack trace" is definitely one grouping option, but there could be more, separate sections. These sections wouldn't need all the table rows but could focus on those that are important to the grouping option to see better at the first glance if they are indeed similar (enough). |
Beta Was this translation helpful? Give feedback.
-
Maybe an option to also manually merge issues will be good? For now we can merge issues only when Sentry mark them as similar. But sometimes Sentry does not see similarity but in out opinion there are the same issue. For example: <?php
$array = null;
echo $array['b];
echo $array['c']; Above example will produce two separate issues ( |
Beta Was this translation helpful? Give feedback.
-
I am indeed very much interested in better issue grouping. In my current project most of the errors are getting picked up incorrectly and thus make it very hard to filter out each error type on its own. They all share the same stack trace, but the messages are different since they come from a 3rd party API. |
Beta Was this translation helpful? Give feedback.
-
Ever so often Sentry notifies you that a new Issue was detected but you’ve already seen this Issue before - you see that this latest Issue has some minor differences in the stacktraces from the original one you viewed but it’s not a new Issue
This problem can get even worse for various Javascript Issues - because there are parameterized URLs that show up in the stacktraces leading to numerous Issues getting created. We’re working on improving this experience for our users.
We are actively seeking users who’d be happy to provide feedback on Issue Grouping as a part of our Beta cohort. (Please reply below if you’re interested or drop us a note at [email protected]
Here’s how we’re solving this problem in 2 phases:
Detect Similar Issues more accurately
When you view an Issue in Sentry, you can go to the new Similar Issues tab and see a list of issues that are very similar based on stacktraces and should be grouped together based according to us:
The “Would group” column represents Sentry’s opinion on whether these issue should have been grouped together instead. You can help determine the accuracy of our improvements by simply agreeing or disagreeing with these similar matchings
This process allows us to identify cases where we’d over-group otherwise causing you to miss genuinely different issues and helps us QA the improvements
Improving the grouping process itself
Once we validate that our enhancements can accurately determine issues that should be grouped together, we’ll start gradually shipping grouping improvements based on these enhancements to help reduce noise for our customers.
We need your help
We are actively seeking users who’d be happy to provide feedback on Issue Grouping as a part of our Beta cohort. (Please reply below if you’re interested or drop us a note at [email protected]
We’re actively looking for feedback on grouping problems you face while using Sentry and we’d appreciate inputs on any of the following:
Beta Was this translation helpful? Give feedback.
All reactions