Diagnose failing checks
Read checker results, run a diagnostic against one box, restart the checker runner, and read filtering verdicts.
When a team reports that its service is marked down unfairly, or every team suddenly fails the same check, work from the scored results to the checker itself.
Read the scored results
Open Checks. Every checker attempt appears with its team, service, operation, status, message and finish time, filterable by team and service. Status and message are exactly what was scored.
Statuses:
- OK: the check passed.
- Down: the service didn’t answer in time, refused the connection or dropped it.
- Faulty: the service answered, but wrongly.
- Flag not found: the service works, but a flag it should hold is gone.
- Cancelled: the check was withdrawn, for example by a pause, its checker image was missing on the host, or an imported checker failed. It counts neither for nor against the team.
A checker of your own that crashes, won’t start or stops answering is recorded as Down, with the error as its message, so every team fails the same check at once.
The same data is available with athena ad check list. Players see only a fixed, redacted summary of their own results, never your checker’s messages.
If every team fails the same operation at once, suspect the checker before the teams.
Run a diagnostic
In Checks → Diagnostics, pick a team and a service and dispatch a check. It runs your real checker image against that team’s box as a service check, and never touches scores, flags or checker state. You can run one in any phase, including before the network opens.
When it finishes, open its output to read the checker’s stdout and stderr. Output is bounded in size and scrubbed of flags, tokens and credentials. Only diagnostics keep this output; scored attempts keep status, message and timing.
From the command line: athena ad check run, then athena ad check bundle for the output.
Restart the checker runner
Checker runner shows whether the checker runner is connected, its backlog, the last heartbeat and last result, and any error. If it is stuck, click Restart runner. Checks in flight when the runner restarts end as cancelled, so the tick in progress may record cancelled attempts; a cancelled check counts against no team. A restart doesn’t clear a backlog that comes from checks taking too long; raise the checker timeout or fix the checker instead.
If the runner reports that a checker image is missing, the event pauses and no team is scored for checks that couldn’t run. Resume once the runner reports healthy again.
Read filtering verdicts
Teams may firewall their boxes, rate-limit and block specific attacks. What they may not do is refuse every other player while answering only the checker. The platform compares, per tick and per declared port of each box, how the box answers players’ connections with how it answers the checker.
A port is flagged only on clear evidence: at least three different player addresses each tried it repeatedly and none was answered, while the checker was answered from at least two addresses. Addresses are counted, not teams, so several players from one team can make up the three. A single answered player connection clears that port for the tick, and only that port. Too little traffic, or incomplete evidence, is never a violation; see Fair checking for the full rule.
A confirmed verdict marks every store of the service that declares the port Down on that box for the tick, so the team earns no SLA and no defence for them. Trust & Safety → Filtering detection lists every store that received one, by tick range, and every finding behind them: the tick, team and port, the verdict, why it was reached, how many player addresses tried and how many checker addresses were answered, and the version of the rule that judged it. Inconclusive findings are listed too, and change no score. The store’s scoreboard history shows a penalty as well, and the final data export keeps every finding, so a verdict can still be explained after the event.
From the command line: athena ad integrity filtering.