CAPACITY_VALIDATION_QUICKSTART.md

Capacity Validation Quickstart

Run this procedure only on the dedicated UnoSim_Test staging host. Candidate profile values are validation inputs and must not be copied into production.

Prepare the host

npm ci
npm run build
npm run build:sandbox
docker info
docker image inspect unosim-sandbox:latest

Check that the working tree is on feature/capacity-validation at the commit being measured, and record host CPU, memory, and available storage outside the repository's generated results directory.

Run the historical baseline first

./scripts/run-capacity-tests.sh

This runs BASELINE (SIMULATION_MAX_CONCURRENT=5, SANDBOX_START_MAX_CONCURRENT=5, SIMULATION_ADMISSION_MAX=25) at bursts 5, 25, and 40, then performs the six-client queue-timeout check. It writes receipts, logs, and metrics beneath the ignored capacity-test-results/ directory. The receipt is tied to the current Git commit and harness contents; rerun the baseline after any commit that changes either.

Candidate staging ladder

After a successful baseline, proceed one profile at a time on the dedicated host:

./scripts/run-capacity-tests.sh --profile R20 --burst 100 --dedicated-host
./scripts/run-capacity-tests.sh --profile R40 --burst 100 --dedicated-host
./scripts/run-capacity-tests.sh --profile R60 --burst 100 --dedicated-host
./scripts/run-capacity-tests.sh --profile R80 --burst 100 --dedicated-host

The burst candidate is separate:

./scripts/run-capacity-tests.sh --profile R80_BURST --burst 180 --dedicated-host

R80_BURST represents 80 running plus up to 100 waiting/admitted requests. Because queued runner acquisition expires after about 60 seconds with SYSTEM_BUSY, a short-run pass does not establish compatibility with long simulation leases. Queue timeout remains a separate policy test.

Do not run candidate loads on a developer machine. Preserve each run's output outside Git and verify cleanup before advancing to the next profile.