Contributor: @gabriel.moise , @Killian-Delarue

Context

Building on the successful integration of the PPX Profiler (see milestone here), the goal now is to visualise profiling outputs and formulate a clear statement on the performance improvement of transitioning to OCaml 5. This will be achieved by:

1. Leveraging Prometheus outputs for dedicated Grafana dashboards as an initial step.
2. Progressing towards OpenTelemetry traces for more granular insights.

Deliverables and tasks:

• Study the current landscape (ETA: 29 nov) :
  • (1-2 days) Gather information on grafazos and how it can be used to generate an initial Grafana dashboard (Reference ongoing Slack thread)
  • (hours) Present a Grafana dashboard MR that uses Prometheus outputs added already by the new profiling outputs : !15785 (merged)
  • (1 day) Merge the base Grafana dashboard
• Add more Prometheus traces (ETA: 13 dec)
  • (hours) Synchronise with the Performance monitoring team to decide on what we want to visualise in the dashboards
  • Found bug with Prometheus monitoring only if the parent is also modified (revert this decision - !15818 (merged))
  • (1 day) Add more Prometheus outputs derived from the profiling data (example MR): !15814 (merged)
  • (days) SYNC with infrastructure team to deploy the new dashboard(s) on an existing monitoring node - discussed with @Killian-Delarue
    • initial job to check that building with profiling enabled is not failing in the CI - !15862 (merged)
    • artifacts MR for profiling node - !15938 (merged)
    • child job to create an octez-node binary with profiling enabled for prometheus - TBD (ongoing by @Killian-Delarue 🙏 thanks a lot) - !15901 (merged) (based on the parent MR: !15896 (merged))
• OpenTelemetry integration (ETA : ?? )
  • ! This part is blocked because of the OCaml 5 workers, so there is no OpenTelemetry work.
  • (hours) SYNC with @picdc to determine the status of the OpenTelemetry work
  • (days) Create a new dashboard when there are OpenTelemetry traces implemented for the profiling outputs
  • (hours) SYNC with infrastructure team again to deploy the new dashboard as well