Distinguish DBnomics errors from provider errors
As we already discussed, the red color of the circles in the dashboard makes us think that there is a problem to fix in our fetchers. But sometimes it is not the case, and it is difficult to operate the platform when real errors and expected (or expectable) errors are presented side by side with the same red circle.
We would like to eliminate or make distinguishable those expected errors, because they are false positive. We don't want to consider that it is an error from our side (e.g. fetcher source code), but on the other hand we want to be aware that the download failed.
I'm looking for a good solution for that problem.
@bduye @pdi @thomasbrand @MichelJuillard Do you have any opinion about how to handle this?
Propositions
- Transform expected errors to success, displaying them as green circles.
- Benefit: we would not be polluted by those false positive errors.
- Downside: this would prevent us to be aware of the failure.
- Introduce a new color for circles (orange? empty circle with border only?) used for expected errors.
- Benefit: we could distinguish expected errors from unexpected ones.
- Downside: too many colors could be difficult to read.
- Add a mark next to red circles meaning "that's an error but there's nothing to do"
- Add filters to the dashboard to view only some "colors"
- Add a state to the dashboard to remember the association between errors and issues, or to mark some errors as resolved and give a reason
Examples
- #574 (closed) there is a problem with the servers of the provider that are sometimes unreachable.
- #593 (closed) some series of some datasets are known to be impossible to download, so they are considered by the fetcher as an exception