Open
Description
I'm still trying to narrow down the exact conditions but the thing that triggered this behaviour was setting the default start in config.yml to a date far in the past.
model_defaults:
dialect: trino
start: 2020-03-09
After doing this we noticed the plan mentioned 4 batches which then failed our duplicate primary key audit as there's no temporal macros used in the full models, so it inserted the same date 4 times.
I initially thought it could be because of the variable BATCH_SIZE = 10000
as our models are set to interval_unit 'hour'
but that only seems to be used when creating source queries for dataframes.
At the moment we have a workaround by changing the start directly in full models to start '1 day ago'
Metadata
Assignees
Labels
No labels
Activity