Best practices
Keeping suites fast
Where time goes in a suite and the SDODS settings that recover it.
What you'll learnThe main costs in a suite and the levers SDODS provides for each.
| Cost | Lever |
|---|---|
| logging in per scenario | auth.storageState: true and @user:<role>; login happens once per pool user |
| browser startup | keep API scenarios @api so they never open a browser |
| slow third-party network | HAR replay with --strict; offline CI |
| screenshots | step policy only for @regression; onlyOnFailure on slow environments |
| locator timeouts | heal contexts resolve in seconds, not minutes; fix the primary locator when a heal repeats |
| serial suites | --workers, --shard, fullyParallel by default; use @mode:serial only where state demands it |
| retries hiding slowness | insights show lists the slowest and flakiest scenarios; fix, do not retry |
Measure with run --json (durations in summary.json) and the dashboard's duration chart.