Summary
There is no UI path to query all transactions where a specific profiled function exceeded a given duration threshold. The backend EAP profileFunctions dataset already stores this data with function, function.duration, and transaction_name as filterable attributes, but the Explore UI has no "Profile Functions" tab to surface it.
Current behavior
The existing profiling UI surfaces function performance in three ways:
- Slowest Functions widget — aggregates function performance across all profiled transactions, but links only to individual flamegraph captures, not a filtered transaction list.
- Suspect Functions table — scoped to a single transaction type, same limitation.
- Function Trends widget — navigates to individual profiles only; no cross-transaction search.
The EAP API is also not publicly exposed, so customers cannot query the profileFunctions dataset directly as a workaround.
Gap
Users cannot answer the question: "Find all transactions where function X exceeded N ms" — even though the backend can answer it. This matters for understanding how widespread a performance regression is across an application, rather than just knowing it was slow in one captured profile.
Verified from source (sentry/profiles/task.py, EAP attributes.py):
- Each function call occurrence is stored as a separate
TraceItem row with function, function.duration, transaction_name, and profile_id.
- Filtering
function = "myFn" AND function.duration > 100ms is a valid query against the dataset.
trace_id is currently set to profile_id as a placeholder, so joining back to the spans/transactions table via trace context is not yet supported.- Profile function ingestion into EAP is behind an internal feature flag and may not be broadly enabled.
Options
-
Add a "Profile Functions" tab to Explore — backed by the profileFunctions EAP dataset, functionally analogous to the existing Spans tab. Users could filter by function, function.duration, transaction, etc. and get a list of matching occurrences with profile_id deep-links. No new backend work required; the data model is already in place.
-
Expose function-level aggregation in the dashboard widget builder — surface profileFunctions as a selectable dataset in dashboards. Lower discovery friction but less flexible than Explore.
-
Add cross-profile search to the existing Slowest Functions / Suspect Functions views — augment current widgets to link out to a filtered list rather than individual flamegraphs. Narrower scope but faster to ship.
Recommendation
Option 1 (Explore tab) is the most composable and directly unblocks the use case without new backend work. It aligns with how Spans exploration already works and sets a foundation for future function-level alerting or dashboarding.
Prerequisite: confirm rollout status of the EAP profile function ingestion flag before committing to a GA timeline.
Action taken on behalf of Prithvi Rajakumar.
Summary
There is no UI path to query all transactions where a specific profiled function exceeded a given duration threshold. The backend EAP
profileFunctionsdataset already stores this data withfunction,function.duration, andtransaction_nameas filterable attributes, but the Explore UI has no "Profile Functions" tab to surface it.Current behavior
The existing profiling UI surfaces function performance in three ways:
The EAP API is also not publicly exposed, so customers cannot query the
profileFunctionsdataset directly as a workaround.Gap
Users cannot answer the question: "Find all transactions where function
XexceededNms" — even though the backend can answer it. This matters for understanding how widespread a performance regression is across an application, rather than just knowing it was slow in one captured profile.Verified from source (
sentry/profiles/task.py, EAPattributes.py):TraceItemrow withfunction,function.duration,transaction_name, andprofile_id.function = "myFn" AND function.duration > 100msis a valid query against the dataset.trace_idis currently set toprofile_idas a placeholder, so joining back to the spans/transactions table via trace context is not yet supported.Options
Add a "Profile Functions" tab to Explore — backed by the
profileFunctionsEAP dataset, functionally analogous to the existing Spans tab. Users could filter byfunction,function.duration,transaction, etc. and get a list of matching occurrences withprofile_iddeep-links. No new backend work required; the data model is already in place.Expose function-level aggregation in the dashboard widget builder — surface
profileFunctionsas a selectable dataset in dashboards. Lower discovery friction but less flexible than Explore.Add cross-profile search to the existing Slowest Functions / Suspect Functions views — augment current widgets to link out to a filtered list rather than individual flamegraphs. Narrower scope but faster to ship.
Recommendation
Option 1 (Explore tab) is the most composable and directly unblocks the use case without new backend work. It aligns with how Spans exploration already works and sets a foundation for future function-level alerting or dashboarding.
Prerequisite: confirm rollout status of the EAP profile function ingestion flag before committing to a GA timeline.
Action taken on behalf of Prithvi Rajakumar.