Closed
Description
Recently we had a collision on "experiment run id" for three automated experiment runs. We should reconsider how we determine this value. Currently it's computed as the hex representation of the current timestamp seconds.
At first it seems seems unlikely that two or more experiments will compute the same run id, but given that we now have 39 (at the time of this writing) projects running automated experiments, each running 2-3 experiments, and them all starting at the exact same time Sunday @ 9:00 UTC, it's bound to happen often.
Metadata
Metadata
Assignees
Labels
No labels