Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
Description
Context
Currently, our error metrics don't distinguish between distribution failures that are permanent and will fail even if retried, or failures that succeed after being retried.
We want to improve this in order to be able to differentiate both scenarios.
Solution
Failure metric should be labeled by:
- Transient failure
- Permanent failure
Proposed approach
We can distinguish both these scenarios by using the following rationale:
- Transient failures happen whenever a package is distributed successfully but had more than 1 attempt at being distributed: retries > 0
Attachments
Issue Links
- links to