On reproduce.d.n I regularly see disconnected workers. I assume this is due to rebuilderd getting restarted daily for a DB backups. The good thing is that rebuilderd still accepts the results, once the worker is done. But sometimes it also schedules the same job again. This happens especially for long running jobs. I see two ways to mitigate this: 1. reconnect disconnected workers. 2. Add some penalty time for jobs with unknown state so they are not scheduled again immediately.
On reproduce.d.n I regularly see disconnected workers. I assume this is due to rebuilderd getting restarted daily for a DB backups. The good thing is that rebuilderd still accepts the results, once the worker is done. But sometimes it also schedules the same job again. This happens especially for long running jobs. I see two ways to mitigate this: 1. reconnect disconnected workers. 2. Add some penalty time for jobs with unknown state so they are not scheduled again immediately.