Pulp

To reproduce¶

1. Start a pulpcore-worker against an empty database
2. Observe that the status API shows that worker
3. kill -9 your worker
4. Observe the status API after 30 seconds no longer shows your worker
5. Start your pulpcore-worker again
6. Go into shell_plus and observe this query shows 2 workers present: Workers.objects.count()

Expected result¶

In [1]: Worker.objects.count()
Out[1]: 1

Solution¶

Have workers that startup run a query that delete any workers that haven't issued a heartbeat in say 7 days. This does not need to be configurable. The 7 day idea is to make something that won't let records accumulate on the long term, but leave them in place for someone to look at the db post-mortem and still see them for investigation.