Open padraic-shafer opened 6 months ago
The SQLite usage here is coming from a databroker backend that was experimental in ~2015 and never used on the floor at NSLS-II (or anywhere else, to my knowledge). There is a strong aversion to deleting any code from this project, but perhaps we could xfail these tests….
Or mark them as flaky? https://docs.pytest.org/en/latest/explanation/flaky.html#plugins
I occasionally see timeouts in unit tests run on GitHub, stemming from the sqlite module. I'm not sure whether we simply need to increase the timeout for these tests, or if there is something more subtle like a race condition between tests using the db.
https://github.com/bluesky/databroker/blob/b7846d5e4d921338bf51fa73fafe4375e1b148ec/databroker/assets/sqlite.py#L121-L123
The failures are infrequent, but could be suggestive of being "on the edge" of an instability that might be triggered by the timing of usage on systems in the wild.