Hi — I have two models deployed in Numinous under the user unpleasant-eduardo (naty-condor, model #16257, and naty-quant, model #16258). Both show Runner: Running, but neither has received a single event for over four days.
What the runner logs show:
naty-condor (deployment #20089, created 2026-07-30 17:50): setup completed fine and it predicted 211 events between 22:55:15 on Jul 30 and 00:11:33 UTC. After that, nothing. The last log line is: Handshake failed with error SSL_ERROR_SSL: error:100000c0:SSL routines:OPENSSL_internal:PEER_DID_NOT_RETURN_A_CERTIFICATE: Invalid certificate verification context (I0801 13:51:08). I confirmed the log viewer was not truncating — I scrolled the container to the bottom and those 211 prediction lines are all there is.
naty-quant (deployment #20172, created 2026-08-02 23:18): setup completed, zero predictions.
What I already tried, following the advice enzo gave in the deployment #19079 MODEL_RUNNER_CONNECTION_FAILED thread (“Just restarting or re-submitting would be enough. Network drop sometime happen”):
1. Stopped and restarted the runner.
2. Re-submitted both notebooks and created fresh deployments: #20177 (naty-quant) and #20179 (naty-condor).
3. Both new deployments reached Runner: Running with clean setup logs (ModelRunner started and ready to serve, tracker class found and instantiated).
4. Still zero events on both, more than an hour after deployment. For comparison, naty-condor received its first event about 90 seconds after its original deployment on Jul 30.
Why I do not think this is a platform-wide outage: I read the leaderboard twice, one hour apart, and other participants Brier scores changed in between (for example cute-air / obnoxious-silverfish geopolitical Brier moved from 0.1142 to 0.1132), so events are clearly still being dispatched and scored for others.
Also: the METRICS tab shows “No data available” for both models on both the Main and Signal tracks for the whole Jul 30 - Aug 3 range, even though the runner logs prove naty-condor did produce 211 predictions on Jul 30-31.
Is there something on my side I am missing (a registration step, a per-account limit), or does the dispatcher need to be reset for this account? Happy to provide deployment IDs, logs or anything else useful. Thanks!
