tsuga-analyze-trace-latency

To find trace operations causing high latency, compute p95 duration per span.name over a time window to rank the slowest operations. This approach highlights peak intervals and separates sustained degradation from transient latency spikes for specific services.

Diagnosing latency spikes involves checking whether p95 latency remains elevated across consecutive 5-minute windows. This distinguishes sustained performance regressions from transient spikes by analyzing trace duration trends over the selected time window.

Correlating ERROR log events with trace latency peaks requires searching logs for errors during the worst latency window. This helps support or refute error-related causes, optionally linking logs and traces via trace_id if both data sources exist.

Yes, you can use tsuga aggregation timeseries to analyze p95 span latency. It computes latency and span counts over time windows, while tsuga services list verifies trace availability to ensure the target service has the required data.

Yes, trace latency analysis works for troubleshooting slow requests across any chosen time window. By computing p95 duration per span.name and selecting peak slow operations, it identifies performance bottlenecks regardless of the analyzed period.