So the packages get uploaded to the testing repository when the current crop of package builds finishes. As it stands, that is the only time it happens. If that fails (as it did in Mrel_sync-packages-to-testing_bionic_amd64 #644 [Jenkins]), then the packages may be deleted and not uploaded. That’s what happened in this case.
So things should now be back to normal, but it does suggest that we may want to re-run that job periodically just in case of a failure. I’ll talk to the infrastructure team about this and see what they have to say.
In the future this would be good for osrf/infrastructure if it’s hosted content that’s problematic that’s a good place for us to track it and we can redistributed/refile it and close it out if necessary.
We could then trace down if it’s a configuration issue or a software issues and fix it in the ros_buildfarm software or in the deployment mechanism or in our specific deployment.
That deployment is programmatically open loop as historically the sync has been very reliable and has not had any documented failure modes that need recovery. We should resolve this one and then it won’t need. However it does send an email report to the distro’s release manager as well as reporting error status on the main dashboard which is periodically checked as well.