Overview

Load agents generate controlled traffic patterns to validate latency, error rates, and saturation behavior under expected and peak volumes. Results feed reliability scoring and capacity planning.

Enable agents per organization in Automation → AI Agents. Human review applies to generated output before release-critical use.

Who should read this

QA engineers, SREs, platform teams, and developers operating Zof Console and APIs.

Prerequisites

Organization administrator approval to enable agent category
Staging or approved test environment reachable from execution plane
Requirement linkage for audit-oriented teams (recommended)

When to use this workflow

Pre-holiday traffic simulation on staging
Validation after autoscaling policy changes
Baseline before major feature launch

Step-by-step procedure

Define load profile

Document target RPS, duration, and ramp shape with SRE.

Use dedicated load staging environment when available.

Enable load agents

Automation → AI Agents → Load → Enable.

Schedule off-peak windows to avoid noisy-neighbor impact.

Review metrics

Operate → Runs → load run → compare latency trends.

Export results for capacity review meetings.

Key concepts

load validation scope: P95/P99 latency under target RPS; Error rate thresholds during ramp-up; Connection pool and queue depth behavior; Graceful degradation when dependencies slow.
Check 1: P95/P99 latency under target RPS
Check 2: Error rate thresholds during ramp-up
Check 3: Connection pool and queue depth behavior

Best practices

Coordinate load tests with infrastructure teams
Never run unsanctioned load against shared staging without notice
Compare results run-over-run using Runs comparison view