Repository navigation
[Observability] Continuous profiling: gateway endpoint and flame-graph panel #5374
Closed as not planned
Description
Feature Summary
The dashboard has no way to view continuous CPU/memory profiles for a workflow, so diagnosing hot paths and resource regressions requires external tooling. This adds scoped profile retrieval and a flame-graph view in the observability dashboard.
Proposed Solution or Design
- Backend: a gRPC-Web client to the Parca server (
ParcaClient) with a profiles query builder andparseProfiles; aProfilesResourceexposing scoped profile queries (optional workflow and execution id), registered inTexeraWebApplication. - Frontend: a profiles panel with a flame-chart component, scoped by workflow and execution id, with explicit empty states for "no samples yet" and for environments where profiling is disabled.
Affected Area
Workflow UI, Other
Activity
Ma77Ball commentedon Jun 5, 2026
/take
Ma77Ball commentedon Jun 5, 2026
/parent-issue #4070
[-][Observability] PR8: profiles end-to-end (gateway + UI panel)[/-][+][Observability] Continuous profiling: gateway endpoint and flame-graph panel[/+]Ma77Ball commentedon Oct 20, 2026
Closing as part of scoping the observability work to instrumentation and export only. Per maintainer feedback on #5378, Texera will export OpenTelemetry data to external tools (Grafana, OpenSearch, Jaeger, etc.) instead of shipping its own query gateway and dashboard UI. The retained work is in #5375, #5376, #5377 (issues #5367, #5368, #5369).