we have a multiplatform project with artifacts generated on windows, Linux and mac.
Both windows and mac runners are on a different remote network, Linux is a docker runner on the gitlab host.
We opened an issue to track the problem Issue 1912, and we found a workaround (upload using the api manually).
What I’d like to know is if someone else have had a similar problem and if not, if some developer can suggest some good places to start debugging the problem.
What I already did is to rise to
300 each timeout related entry in the conf file, but the fact I can upload using an external tool seems to put the spotlight on the runner side.
The fact a failed upload exits with
0 is already marked in the code artifacts_uploader.go#L72.