Hi everyone!
We have a Java application which sends events to Sentry by default SDK.
We have some problems with merging the same events by stack trace.
In the documentation, I read that events should be merged by stack trace firstly.
But in some random cases (we didn’t find any correlation), it isn’t. Can someone say why does it happen?
By any change is this sentry.io? If so, could you provide a link to the two events so we can investigate?
No, it’s Sentry 9.1.2 as open source software.
Does it have any bugs like this?
We also noticed that there are 2 issues with the same stackrace - in one there is only one event, and in the second all the others. Moreover, the first event in both issues appeared at the same time (accurate to the second). Could it be linked with the problem?
A lot has changed since Sentry 9, I’m not sure if there was a bug exactly for this case.
It’s possible that two events were ingested so close together that their group calculation happened concurrently and somehow they ended up in two different groups.
If it’s exactly the same stacktrace, it should be grouped together. Now if they have almost the same timestamp, that would explain why they were not.
You can group them together through the user interface as a last resort
Thank you.
We tried to reproduce the case locally. It isn’t.
Two same events with timestamps "received":1594289694.116 and "received":1594289694.109 were merged into one issue (both have "datetime":"2020-07-09T10:14:53.000000Z")
In the case above timestamps were "received":1594020958.734 and "received":1594020958.952 (both have "datetime":"2020-07-06T07:35:58.000000Z")
You can look at the ‘hash’ or calculated ‘fingerprint’ values, though I’m guessing we dont expose those as easily in 9.x.
Generally speaking, most errors are hashed via (errorType, specificFragmentsOfEveryFrame). What Bruno said is possible but its more likely that there is a slight variance in the error.