Hi,
I used to just comment the EXTRA_VARS entry in the env_vars_to_src.sh file as it seems to always leads to an error.
The problem here is that we dlrn_hash_tag to the last value found on the undercloud. So in a rocky to master upgrade that would be the master hash, is that case 965941f1e62cef16967e7a7cd6d98263e52acb62_0989b280
But when we deploy, we use rocky and we end up with:
2018-10-24 09:18:41 | ImageNotFoundException: Not found image: docker://docker.io/tripleorocky/centos-binary-cinder-api:965941f1e62cef16967e7a7cd6d98263e52acb62_0989b280
2018-10-24 09:18:41 | time="2018-10-24T09:18:41Z" level=fatal msg="Error reading manifest 965941f1e62cef16967e7a7cd6d98263e52acb62_0989b280 in docker.io/tripleorocky/centos-binary-cinder-api: manifest unknown: manifest unknown"
Looking https://hub.docker.com/r/tripleorocky/centos-binary-cinder-api/tags/ we can see that 965941f1e62cef16967e7a7cd6d98263e52acb62_0989b280 doesn't belong there but there https://hub.docker.com/r/tripleomaster/centos-binary-cinder-api/tags/
The workaround is just to comment the EXTRA_VARS line in the env_vars_to_src.sh file before loading it.
The proper solution would be to save somehow the release.sh file so that it get regenerated in the reproducer node so that we get the same hashes during the reproducer run.
