Transaction tabs

When you open a transaction type in the UI, the detail view is split into tabs. Most tabs show aggregates over all traffic for that transaction; only Traces is limited to slow requests (see Transaction configuration).

To interpret one stored request (Breakdown, Entries, Queries, Service Calls, Profile), read How to read a Trace. For a fix workflow, see Re-engineering a slow call.

Response-time averages and breakdown timers for the selected time range. Useful for spotting which sub-timers dominate typical latency.

p50, p95, p99, and related percentiles for overall response time and breakdown timers. Compare with Average when the distribution is skewed by outliers.

Transaction count and rate over time. Counts reflect all invocations, not only traced ones.

Individual slow requests stored because they exceeded the slow threshold. If nothing appears here, either nothing was slow enough or tracing is disabled — check Transaction configuration. Trace list respects retention and agent connectivity.

JDBC/database activity captured by Glowroot's JDBC weaves (Plugins). Custom Instrumentation does not fill this tab by itself.

Outbound HTTP-ish calls instrumented by Glowroot Plugins (e.g. Apache HttpClient, OkHttp, RestTemplate). This is not an inventory of local @Service or in-process Spring beans — only external/service-boundary calls the plugins recognize (What Glowroot does not do · issue #746).

• Transaction-level Service Calls (this page’s aggregate view) summarize outbound activity across the selected time range — good for trends.
• Inside a single Trace (How to read a Trace), Service Calls answer: for this request, how many outbound calls, to where, and whether the same destination repeats (fan-out / HTTP N+1).

When re-engineering a call, prefer the per-Trace view for evidence, then confirm the fix with aggregates / percentiles.

CPU stack samples collected during traced (slow) requests. Use this when breakdown timers do not sum to the root transaction time — gaps often mean time outside instrumented timers (GC, contention, uninstrumented code). Profiling interval is configured under Transaction configuration.

max trace entries (agent config) can truncate the Entries tree in the UI while aggregate counts on other tabs still reflect the full trace. If the flame tree or entry list looks cut off, raise the limit or narrow the slow threshold so fewer huge traces are stored.

• How to read a Trace — anatomy of one Trace
• Re-engineering a slow call — playbook
• Transaction configuration — slow threshold, profiling, thread stats
• Plugins — what Service Calls and JDBC capture
• Instrumentation — custom timers vs Queries tab
• Plugin coverage gaps
• What Glowroot does not do — scope limits (no full code coverage, no arbitrary log tailing)