You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We want to perform random sample tracing of the Boost RPC api to be able to understand the performance of RPC calls and report those to a local Grafana instance (consider using Tempo for this, for future integrations mention below). Emphasis will be on retrieval for our immediate goals, but assuming sampling of the whole api is an easier top level integration, adding tracing across all calls would also be useful.
Requirements
Add support to the RPC API for enabling tracing, with a configurable sampling rate (ie: 0.1).
This should be disabled by default. We'll likely want to pass a flag to the daemon to enable/set this.
Report traces to a local instance of Grafana
[Nice to have] spin this up as a container addition to docker support
Future work
Support parent trace context sharing across Boost services, such as booster bitswap passing its trace ID to the RPC API, so that we can trace e2e calls.
There is work being done that's nearing completion to add docker support to Boost for testing, adding an instance of Grafana to this for testing could be useful for development on this issue.
The text was updated successfully, but these errors were encountered:
Summary
We want to perform random sample tracing of the Boost RPC api to be able to understand the performance of RPC calls and report those to a local Grafana instance (consider using Tempo for this, for future integrations mention below). Emphasis will be on retrieval for our immediate goals, but assuming sampling of the whole api is an easier top level integration, adding tracing across all calls would also be useful.
Requirements
Future work
References
The text was updated successfully, but these errors were encountered: