Logs with a FAILURE status are currently filtered out of the query.
We need to figure out a way to mark how many attempts have been made for a particular delivery id preferably without causing an update to the row in BigQuery.
The simplest approach would be to just not write a FAILURE status and let the query try again.
The drawback is that something could be stuck forever in that situation and we wouldn't want to keep processing it. I think a secondary FAILURE table might work that we can join into the main query as WHERE count(failures where delivery_id = x) < 10 or something.
This adds complexity to the write operation though so it requires some thought.
TL;DR
Logs with a FAILURE status are currently filtered out of the query.
We need to figure out a way to mark how many attempts have been made for a particular delivery id preferably without causing an update to the row in BigQuery.
The simplest approach would be to just not write a FAILURE status and let the query try again.
The drawback is that something could be stuck forever in that situation and we wouldn't want to keep processing it. I think a secondary FAILURE table might work that we can join into the main query as WHERE count(failures where delivery_id = x) < 10 or something.
This adds complexity to the write operation though so it requires some thought.
Detailed design
No response
Alternatives considered
No response
Additional information
No response