Container Stability Test

A performance test environment must itself be stable. An unstable base - one that produces wildly different results run to run - makes it impossible to tell whether a change in numbers reflects a real regression or just noise. Any signal suggesting intermittent dips or spikes in response times is therefore a concern, even if the absolute numbers look acceptable. The CV (Coefficient of Variation) is the primary measure of this stability: it expresses how much p95 varies across runs as a percentage. A low CV means the environment is behaving consistently and results can be trusted; a high CV means the opposite - investigate before drawing any conclusions.

Each row in the table below shows response-time statistics for one URL path, measured by k6 during this pipeline run. Times are in milliseconds.

• p95 is the primary signal: 95% of requests completed within this time. It is more representative than the median (which hides tail latency) and more stable than the max (which reflects one-off spikes).
• Red p95 means the value exceeded the configured threshold - the environment is not meeting the expected stability target.
• CV (Coefficient of Variation) measures how consistent p95 is across all historical runs. Under 8% is stable; 8–15% suggests noisy conditions; over 15% means results are too variable to be trusted - check runner load or container startup behaviour.
• Sparkline shows the p95 trend across up to the last 50 runs. A flat line indicates a stable environment; a sustained shift up or down suggests the environment has changed; a jagged pattern indicates instability.

The Run History table below shows the mean of all page p95s per run - a single number to scan for environmental change across runs. Click a timestamp to view the full report for that run.