[APM] (Meta) Backends UI: New experience for analyzing uninstrumented backends and services.

[formgeist]

Summary

Original brief statement

Today APM agents collect details about any external calls made from the instrumented service, like HTTP, SQL, Cache, PubSub systems, etc. This includes edge metrics as aggregate statistics and spans with details.

Some of the external calls resolve into a downstream service that is instrumented, and therefore performance troubleshooting is possible by leveraging distributed tracing and drill down into those downstream services.

In case downstream service is not instrumented or not possible to instrument (like Database, or third-party service) - troubleshooting performance issues related to those services isn't easy to be done at scale today.
Historically more than 50% of the problems occur due to a problem introduced by a dependency.

This new page would be accessible as a drill-down straight from APM Service Map and from the APM service overview page when navigating service dependencies.

7.15 scope

We will commit to the following pages and components;

• New landing page for a backend
• New global list pages for all backends
• New service view for dependencies
• Links to the selected backend landing page from a service map backend node, the service overview dependencies table, and finally the trace timeline span detail flyout.

What is not in scope

• The operations breakdown per selected backend (relies on span name metrics which is targeted for 7.15)
• The operation detail page (relies on span name metrics which is targeted for 7.15)

Issues

• [APM] Backends UI: Backend landing page #103253
• [APM] Backend UI: Global backends list page #103256
• [APM] Backends UI: Service view for all dependencies #103257
• [APM] Backends UI: Add link from Service Map backend nodes #103259
• [APM] Backends UI: Add link from the Service overview dependencies table #103260
• [APM] Backends UI: Add link from the span detail flyout in the trace timeline #103264
• [APM] Backends UI: Implement telemetry tracking #103917

Solution

The design issue has a more detailed presentation of the overall concept.

Please note that these screens were put together before we changed the overall page template and navigation, but the contents don't change;

Backend landing page

The backend page will feature an overall performance health view of the backend based on the metrics from all callees. The list of upstream services breaks down the health of the callee services.

The operations breakdown and analysis is not in scope for 7.15 as we're dependent on changes in the agents and server to get span metrics.

Global list of backends

We will also add a global list of backends alongside the other global views like Services, Traces, and Service Map.

Service dependencies tab view

Each selected service will also feature a new view for inspecting the dependencies, which duplicates the chart and table from the Service overview into a nice full-page overview.

Telemetry

Opened #103917 to track telemetry implementation.

Resources & links

Design issue