Closed MattTriano closed 1 year ago
That strategy worked. I expect I'll have cause to revisit this at some point (when another upstream data entry error causes another DAG failure), but I'll come back to that when I have that data point.
Post script: By the way, the source table for this issue only keeps the last 90 days or 3 months of data, so that's why the offending record fell off. So I was inaccurate earlier when I ascribed this to the source table's dtype, when in reality it's just the dtype pandas infers when reading the data into the database.
The dtype of
inventory_number
in the source table forchicago_towed_vehicles
changed fromtext
tobigint
. I can manually alter the column dtype via the following steps.Decide on remediation:
I don't really care about completeness on this table; I mainly included it because it updates all the time and the dataset is small, which makes it convenient for testing things in development. So I'll manually drop the entire record (and be cranky next time a data entry error causes an error when dbt tries to UNION new data with the existing data).
I'll have to think about strategies for handling this more durably.