[TASK] eSignet Go Performance: Confirm OTLP observability config for load runs (disabled or sampled, reachable collector)

[rachik-hue]

Overview

Observability/tracing is hard-enabled by default, which is fine for normal operation but risky for a load test: if the OTLP collector endpoint isn't reachable from the perf environment, or sampling isn't turned down, the exporter will retry in the background during the run — adding overhead and noise that skews the very performance numbers the test is trying to measure.

Description

Files: perf-env deployment.yaml (observability: block)

Actual:

Observability.Enabled is hard-set true in applyDefaults, with no check on whether the perf environment actually has a reachable OTLP collector or a sane sample rate configured.

Expected:

• Confirm the OTLP sink for the perf environment is either explicitly disabled, or pointed at a real, reachable collector.
• If enabled, set sample_rate ≤ 0.01 for load runs, so tracing overhead stays negligible at 50 TPS.
• Verify that an unreachable OTLP endpoint does not cause background exporter retries during the test — this silently adds load and can distort latency/throughput results.

Acceptance Criteria

• Perf environment's observability: block is confirmed to have either Enabled: false, or Enabled: true with a reachable collector endpoint and sample_rate ≤ 0.01.
• During a load run, no OTLP exporter connection errors or retry log spam appear.
• Load test results (latency, throughput) are unaffected by tracing overhead — confirmed by comparing a run with observability on vs. off, if feasible.

Definition of Done

• Perf-env deployment.yaml observability: block reviewed and set correctly (disabled or sampled + reachable collector)
• sample_rate set to ≤ 0.01 if observability is enabled for load runs
• Verified no exporter retry/connection-error noise during a load run
• Documented the chosen configuration and rationale for future perf runs
• Code/config reviewed and merged into develop-go

[Md-Humair-KK]

#2514