mirror of
https://github.com/google/nomulus
synced 2026-01-06 21:47:31 +00:00
ff39a4a763
For reasons that I cannot explain, the same expand recurring billing event pipeline would fail in us-east1 but succeed in us-central1. See: https://pantheon.corp.google.com/dataflow/jobs/us-central1/2023-02-09_14_52_24-162498476138221714;graphView=0?project=domain-registry https://pantheon.corp.google.com/dataflow/jobs/us-east1/2023-02-09_14_26_07-4564782062878841960;graphView=1?project=domain-registry Also improved how the accuracy of the metrics: It is observed that both counters are consistently higher for the same start and end times when running in dry run mode. There is no way to test for consistency when not running in dry run, for obviously reasons. I can make the recurrings in scope counter consistent by not updating it in a side-effect-causing transaction, but there is no way around the other counter. It can only be trusted when running in dry run mode, unfortunately.