sharp dagger
Hi,
After monitoring it for a few weeks, it seems that the "AI enhanced grouping" for us is doing more harm than good (as it merges issues way to agressively, even when the stacks and messages are clearly different).
Is there a way to disable it completely on a project ?
We know that we could change the fingerprint to "enforce" at least the message and the calling function, so it would prevent most of these bad merges, but it would also cause pain for a while (lots of duplicates).
chilly jacinth
Hi @sharp dagger, can you share some example issues with me at [email protected]? This grouping enhancement is designed to be extremely conservative in grouping (to prevent over-grouping) so I'd like to investigate further and see what we can do.
sharp dagger
Hey Matt,
Sure, i'll send you some tommorow morning 🙂
do you have / need access to our sentry instance or are screenshots enough ?
chilly jacinth
If you can send me the urls to the issues I can look up the grouping details directly.
main narwhal
Any resolution to this one?
We are facing the same problem with errors getting grouped, despite clearly very different stack traces.
In the Event Grouping Information section, it lists more than 25 similar frames, which is maybe the problem, but I don't see any way to manage this. We are also aware that we could maybe work around it with custom fingerprint rules as OP suggests, but it seems like bad direction to go.
sharp dagger
Hey 🙂 We're still facing the same issue from time to time. Our QA took the habit to split them manually, but it's still a pain
main narwhal
Thanks for the update, even if it wasn't the result I was hoping for ><
By manually splitting, do you mean using the "unmerge" functionality, or are there better ways to do it?
chilly jacinth
@main narwhal if you can DM me the urls to some events which you wouldn't have expected to be merged I'll have the team take a look