Choosing Between Virtual Users and Arrival Rate in k6 for Throughput Control
Learn when to use Virtual Users versus Arrival Rate executors in k6 to avoid coordinated omission and ensure stable throughput during performance benchmarks.
Loading…
ReadMeFeed / Community knowledge
#load-testing
Real questions. Useful conversations. Find the people who know your stack.
k6Parameterizing Gatling Simulations with CSV Feeders
Learn how to use Gatling CSV Feeders to parameterize HTTP requests, avoid server-side caching, and choose between random, circular, and queue distribution strategies.
gatlingSetting Latency Pass/Fail Criteria with k6 Thresholds
Learn how to use k6 thresholds to turn HTTP latency measurements into clear pass/fail signals for your load tests and CI pipelines.
Scale Your Load Tests on Kubernetes with Karat Labs – A Practical Guide
Scale load tests automatically with Karat Labs. Learn how to define JSON scenarios, let Kubernetes autoscale workers, monitor real‑time metrics, and embed tests in CI pipelines.
karatelabs
kubernetesGatling 3.10 changed the httpProtocol caching default — do upgraded simulations need explicit disableCaching?
I'm reviewing a Gatling simulation suite that was written against 3.9 and recently upgraded to 3.10. The tests hit state-dependent REST endpoints where each virtual user presents a session-specific token, and the payloads include timestamps that must be fresh on every iteration. My understanding is that before 3.10, the HTTP protocol cache was enabled by def
Threshold Evaluation Timing in k6 Load Tests
k6 Threshold Evaluation Timing The precise goal is to configure k6 load tests to reliably signal performance regressions without generating excessive alert noise from transient metric spikes. Constraints and Uncertainty The current uncertainty involves balancing immediate failure detection against the need to filter out short-lived anomalies using abortOnFai
HTTP 500 error during POST request in Gatling scenario
Goal Implement retry logic for a POST request that may return an HTTP 500, while avoiding duplicate writes to the target API. Constraints and Uncertainty Gatling does not ship a built‑in retry directive; users typically build retries using exec , when , and session variable checks. Without an idempotency key, a retry can cause the same POST to be processed t
Configuring Sanity Check Latency Under Load
Performance Measurement Constraint The goal is to establish a reliable baseline for latency when validating system sanity under defined load conditions. Uncertainty and Scope The constraint is maintaining a P95 latency below 50ms during a sustained 100 concurrent connection test using the documented sanity check feature. Which documented approach best handle
Simulation DSL source compatibility ends at the Gatling 2.x to 3.x boundary
Gatling 3.x is not source-compatible with the 2.x simulation DSL, which turns a move between the two major versions into a manual migration rather than a drop-in upgrade. HTTP protocol setup, request chaining, and injection profiles written for 2.x do not compile unchanged under 3.x. Since simulations are compiled Scala classes that the runner loads from the
scala