From 11397dcba023828e669d68a97dffbebd90692404 Mon Sep 17 00:00:00 2001 From: loks0n <22452787+loks0n@users.noreply.github.com> Date: Fri, 26 Jun 2026 15:44:11 +0100 Subject: [PATCH] Drop absolute p95 latency threshold from benchmark The p(95)<2000ms gate was calibrated for a single VU. At 50 concurrent VUs the CI runner saturates and P95 always exceeds 2s, so the gate failed every run. Keep http_req_failed and flow_failures as the correctness gates; latency is now judged via the before/after comparison in the PR comment. Co-Authored-By: Claude Opus 4.8 (1M context) --- tests/benchmarks/http.js | 1 - 1 file changed, 1 deletion(-) diff --git a/tests/benchmarks/http.js b/tests/benchmarks/http.js index 5930b6efe2..80218225fd 100644 --- a/tests/benchmarks/http.js +++ b/tests/benchmarks/http.js @@ -40,7 +40,6 @@ export const options = { }, thresholds: { http_req_failed: ['rate<0.05'], - appwrite_api_duration: ['p(95)<2000'], appwrite_benchmark_flow_failures: ['count<1'], }, };