-
-
Notifications
You must be signed in to change notification settings - Fork 4.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Transaction Duration Metrics don't match Measurement Metrics #68747
Comments
Assigning to @getsentry/support for routing ⏲️ |
Routing to @getsentry/product-owners-performance for triage ⏲️ |
Sorry, but the durations were obscured by the excel sheet and they matched. They were listed in tables on the bottom with the only filter on the left being |
Hmm those look correct, you should be getting the same percentile for the same period, although there can be some small differences since we use an approximate quantile function so when there are millions (or billions) of data points it doesn't take a long time. Can you possibly share the id for this dashboard ( |
Sorry for the delay. Let me make sure I can disclose that |
ID: 85476 |
Routing to @getsentry/product-owners-ddm for triage ⏲️ |
Alternatively, if you just want to get the immediate solution you can rename the measurement too which should work immediately. As well yes the unit determination is based on the currently selected projects |
@wmak is right, we can't delete measurements. The best bet would be to use a different name if you'd like to start fresh. slightly unrelated side note: we don't support deletion in custom metrics either; we just support "disabling" which blocks ingestion and hides the metric from UI, but unfortunately, that's not available with custom measurements |
Ok, we'll explore versioning our metric/measurement names. Looking forward to the fix, too. edit: I just wanted to be sure I understand. If we send a measurement to Project1 and then to Project2, is that considered two different measurements? We do have a smaller project just for experimentation and non-production data. |
Not really, its more that per request we look at the selected projects and try our best to figure out what the user means when you request a measurement. If both projects are selected, the unit from either project may be used. |
Routing to @getsentry/product-owners-ddm for triage ⏲️ |
Hey all, any updates here? |
Routing to @getsentry/product-owners-dashboards for triage ⏲️ |
Hi @kylannjohnson, We don’t have an ETA yet for the longterm fix to handle and merge multiple units on the backend. This is currently on our backlog, but if you're still having issues, we'll try to find a solution. Thanks! |
Hey @ale-cota , Thanks for the response. Yes, we have decided to version our measurement names going forward. It seems like this will address our issue in our local tests on a separate dashboard. Perhaps there's already a way to do this, but some kind of Measurement Summary could be a way to indicate what raw values came in to Sentry or what unit they are being assigned. I can see why the implementation is "sticky", but versioning a measurement name also isn't great. I'm sure there's a great solution to this product level issue, but we'll keep an eye out as we grow our usage of measurements in the near future. |
Thanks for confirming that versioning for the names at least addresses the issue for now. We'll update you once we manage to schedule the investigation work for merging/summarizing multiple units. |
Environment
SaaS (https://sentry.io/)
Steps to Reproduce
Consider two transactions, one un-scoped and the other scoped. (Let's call them Transaction A and B). Transaction A's duration is added as a measurement to Transaction B because we'd like to just track the duration of this operation instead of using a transaction on it.
The metrics don't seem to match and my question is whether they should or not.
In the dashboard below the left panels are showing various functions on
transaction.duration
of Transaction A. The lowest panel is just a list of the duration. The panels on the right are showing the same functions but applied to a measurement sent with Transaction B (the scoped one).Expected Result
I expect there's some interpolation going on in both cases, but I would expect the same results, given the same, small set of data. Should a percentile calculation done on a measurement be the same as a duration of the same amount? All durations are known and seem to both be in the data set.
If there is some variance, what is the accuracy of it? e.g. +/- 1%
Actual Result
Product Area
Dashboards
Link
No response
DSN
No response
Version
No response
┆Issue is synchronized with this Jira Improvement by Unito
The text was updated successfully, but these errors were encountered: