Gatling feeder data lifecycle versus injection profile duration alignment

Answer first

Gatling 3.x does not automatically cycle feeder data because the injection profile is long. Cycling, exhaustion or replacement is determined by the feeder strategy you select, not by the injection profile duration. The feeder instance is shared across virtual users and accessed via a thread-safe iterator with an atomic pointer. That prevents iterator corruption, skipped reads and duplicate pointer moves, but it does not enforce logical uniqueness of data values across users.

Confirmed facts

• Strategy-driven lifecycle. circular wraps around indefinitely, queue consumes each record once then exhausts, random picks with replacement, shuffle randomizes once then behaves circular. Lifecycle is set by the strategy, not by ramp-up or simulation length.
• Thread safety. Feeder instances are shared. Access is coordinated so concurrent virtual users cannot corrupt the iterator. Logical duplication is still possible under circular, random or shuffle.
• Materialization impact. Memory materialized feeders load the full data set into heap for fast access. Streamed feeders read lazily from disk, reducing heap pressure at the cost of I/O. Under sustained high RPS, streamed I/O can affect pacing accuracy; large memory feeders can increase GC pauses that affect injection timing.

Likely behavior in practice

In Gatling 3.x the common built-in feeder builders are commonly used with circular semantics, so when total requests from the injection profile exceed feeder cardinality the data is reused via modulo indexing. This is version-sensitive. Defaults can change between minor releases, so verify the explicit strategy on your feeder definition.

Steps needed for this case

1. Inspect the Simulation code for the feeder builder and confirm the presence of .circular, .queue, .random or .shuffle modifier.
2. If at-most-once consumption is required, use queue and ensure record count >= total expected requests from the injection profile. If indefinite reuse is acceptable, use circular or shuffle and accept duplication.
3. Run a minimal simulation with a tiny feeder and an injection profile longer than the feeder size, logging each virtual user’s fed values to observe wrap-around or exhaustion.
4. Check the Gatling 3.x minor version in use and compare its release notes for default feeder strategy changes relevant to your build.

One diagnostic detail that changes the recommendation: which feeder strategy is currently applied in your Simulation, and the Gatling 3.x minor version you are running. Without that, assumptions about default circular behavior remain unverified.