Policy Cheatsheets
Quick starting points for Chain Engine and Rotation Engine policy fields.
Policies are where runtime behavior becomes predictable. The easiest way to tune them is to start from Balanced, change one thing, run traffic, and read the runtime page. If you change timeouts, retry budget, fallback, session limits, and quarantine all at once, the traffic log becomes much harder to explain.
Chain Policies
Chain policies answer a simple question: how should a saved chain behave when it becomes a local proxy endpoint? Startup validation decides whether the route must prove itself before startup is considered healthy. Timeouts control how long the runtime waits. The health interval and failure threshold control when routes are treated as unhealthy. Retry budget and fallback decide how much the runtime should try alternate routes. Optional local proxy auth protects the endpoint for clients that should not use it anonymously.
| Goal | Starting point |
|---|---|
| General chain runtime | Keep Balanced defaults and watch traffic first. |
| Faster failure detection | Lower request/connect timeouts gradually; keep failure threshold moderate. |
| More tolerant chain | Increase request timeout before raising retry budget. |
| Target-specific validation | Use a probe URL and expected status list that reflect the actual target. |
| Local client gate | Enable require auth and set a username/password. |
| Hard traffic limit | Set the bandwidth limit in megabytes. |
Rotation Policies
Rotation policies answer a different question: when should the runtime keep a proxy, switch proxies, cool a proxy down, or quarantine it? Session limits control how long a proxy assignment can live. Rotate-on statuses react to target responses such as blocks or gateway failures. Cooldown gives a proxy a short rest after failure. Quarantine removes repeated failures for longer. Per-proxy concurrency prevents one proxy from taking too much simultaneous load.
| Goal | Starting point |
|---|---|
| Rotate often | Keep max requests per session low. |
| Keep target sessions stable | Enable sticky by target host and raise session limits. |
| Reduce proxy pressure | Lower max concurrent requests per proxy. |
| React to target blocks | Keep block-like statuses in rotate-on statuses. |
| Avoid repeated bad proxies | Increase quarantine seconds, then rebuild the pool if it stays high. |
| Test a new pool quickly | Use shorter cooldown while testing, then return to safer values. |
How To Tune Without Guessing
Watch the runtime counters after every policy change. In Chain Engine, route count, healthy route count, failure count, switch count, and last error tell you whether the chain or fallback path is the problem. In Rotation Engine, available, cooling, and quarantined counts tell you whether the policy is too aggressive or the source pool is weak.
Policy tuning cannot rescue a bad pool forever. If the runtime keeps cooling or quarantining proxies, rebuild the source pool from fresh checker and benchmark results.