Allows individual schedules to run at a fixed timezone (overridding whatever timezone has been set on the instance).
Details
BK
Diff Detail
- Repository
- R1 dagster
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
python_modules/dagster/dagster/core/definitions/decorators/schedule.py | ||
---|---|---|
513–517 | Won't this be confusing in the playground and backfill UIs now? I feel like if they want their schedule to run every hour exactly, they should use UTC execution_timezone. But if they configure an hourly schedule in a DST timezone, the partitions should be local to their timezone and the behavior at DST boundaries should be documented. |
python_modules/dagster/dagster/core/definitions/decorators/schedule.py | ||
---|---|---|
513–517 | Made the partition names (and values) timezone-aware in https://dagster.phacility.com/D4761 |
Ditch the instance-level default timezone (going to tackle getting the partition names and values exactly right separately)
python_modules/dagster/dagster/core/definitions/decorators/schedule.py | ||
---|---|---|
37 | Maybe mark this argument on all the decorators with the new experimental feature flags | |
python_modules/dagster/dagster/scheduler/scheduler.py | ||
213–216 | Where do these show up again? | |
python_modules/dagster/dagster_tests/scheduler_tests/test_scheduler_timezones.py | ||
112–114 | Shouldn't all these tests now freeze the time in UTC? | |
146 | The result would be the same even if this were 1 minute later right? |
python_modules/dagster/dagster/scheduler/scheduler.py | ||
---|---|---|
213–216 | just as output in the scheduler process For 0.10.0 IMO we should require execution_timezone on all schedules, so this should become an invariant violation | |
python_modules/dagster/dagster_tests/scheduler_tests/test_scheduler_timezones.py | ||
112–114 | I don't think so no, there's no guarantee that the server running them will be in UTC | |
146 | yeah, will update to make more clear |
python_modules/dagster/dagster/core/definitions/schedule.py | ||
---|---|---|
149 | Is there some sort of built in type in datetime that we can use here to represent the timezone instead of a string? |
fix some issues with calculating the partition times during DST transitions (can't trust relativedelta or timedelta, can trust pendulum.datetime.subtract and pendulum.range)