Troubleshooting Recipes
Symptom-based fixes for extraction, checking, benchmarking, leak tests, runtimes, pools, and exports.
Troubleshooting is faster when you fix the earliest broken stage. Do not tune Rotation Engine policy because the extractor gave you stale proxies. Do not rewrite a JSON path because the checker target has the wrong expected status. Walk the workflow in order: source, task settings, result filtering, pool state, runtime policy, then client setup.
Extractor Returns Zero Proxies
First prove the source shape. Run Auto once, then force Plain text, HTML, or JSON based on what the source actually looks like. For HTML, start with broader selectors like table tbody tr, pre, or [data-proxy]. For JSON, remove the path once to confirm the payload shape, then add a tighter path again.
If the source sometimes works and sometimes fails, increase timeout and lower concurrency for the test run. If the source consistently returns many candidates but few proxies, the parser is probably seeing too much unrelated text.
Checker Marks Too Many Rows Dead
Dead rows are not always a checker problem. Increase timeout once, use Auto protocol for mixed lists, and make sure the expected status list matches the request URL. If failures cluster by source or provider, the source may simply be stale.
When many rows fail together, lower concurrency and test a smaller sample. If the smaller sample improves, the original run may have been too aggressive for the source quality or local network.
Benchmark Looks Unstable
Benchmark only rows that recently passed Checker. A benchmark full of dead or weak proxies produces noise, not insight. Start with 3 rounds, lower concurrency if results look chaotic, and compare average latency with worst latency. The best single latency is not enough if the worst latency is terrible or the success rate is weak.
Leak Test Cannot Classify
Leak testing depends on the target returning inspectable JSON with headers and origin. If the target does not return that shape, classification will be weak. Confirm the proxy still checks alive, reuse the detected protocol from Checker if Auto is inconsistent, and inspect one surprising candidate in Proxy Viewer.
Rotator Has No Available Proxies
Open the runtime snapshot and read available, cooling, and quarantined counts. If most proxies are cooling or quarantined, the policy is reacting to failures. Re-check and benchmark the source pool before loosening policy. If the source pool is good but concurrency is high, lower max concurrent requests per proxy or increase source size.
Runtime Bound Port Does Not Match
The requested port is a preference. The bound port is the truth. Copy the endpoint from the runtime snapshot and update the client to use 127.0.0.1:<bound-port>. Then send one small request and confirm the traffic page increments request count.
Discord Export Fails
Check the webhook URL, file count, and payload size. Then reduce the export set and try again. Exporting a smaller filtered set is usually better than trying to send a huge raw task report.
If you are stuck, capture the exact stage that first goes wrong: extractor metrics, checker error cluster, benchmark success rate, leak signal, pool health, runtime reason, or client connection failure.