Motivation:¶

Repositories became typed, therefore for each incoming repo we need to identify it's type in order to migrate it
https://github.com/pulp/pulp-2to3-migration/blob/master/pulp_2to3_migration/app/migration.py#L68

for pulp2repo in pulp2repos_qs:
                pulp3_repo_name = pulp2repo.pulp2_repo_id
                repository_class = PLUGIN_MIGRATORS.get(pulp2repo.type).pulp3_repository
                repo, created = repository_class.objects.get_or_create(
                    name=pulp3_repo_name,
                    description=pulp2repo.pulp2_description)

Identifying `repository_class` for each repo might have performance impact

In case we'd like to 'migrate' a repo to pulp3 that is not present in pulp2, or in other words - migrate pulp2 repo with new name in pulp3 we'd have troubles to find its `repository_class`

for pulp3_repo_name in repos_to_create:
                try:
                    pulp2repo = pulp2repos_qs.get(pulp2_repo_id=pulp3_repo_name)
                except Pulp2Repository.DoesNotExist:
                    description = pulp3_repo_name
                else:
                    description = pulp2repo.pulp2_description

                repository_class = PLUGIN_MIGRATORS.get(pulp2repo.type).pulp3_repository
                repo, created = repository_class.objects.get_or_create(
                    name=pulp3_repo_name,
                    description=description)

https://github.com/pulp/pulp-2to3-migration/blob/master/pulp_2to3_migration/app/migration.py#L91

Solution¶

Having repos grouped by plugin type( iso, docker. rpm) would help to deal with above mentioned cases.

Update plan.get_pulp3_repository_setup() so it would have following structure:

{iso: {repoA : repo_data, repoB: repo_data}, docker: {repoA : repo_data, repoB: repo_data}}