Update OAuth App to Use The RemoteRepositoryRelation Model

[saadmk11]

This does not take into account for duplicate RemoteRepository for now.

[humitos]

Changes look GOOD! I left some comments to keep talking, but there isn't anything too big to change.

This does not take into account for duplicate RemoteRepository for now.

Is this comment old now? The code seems to handle duplicated RemoteRepository properly now.

[humitos]

It may be good to start populating remote_id here as well. I think we already added to the modeling.

It will be easier to populate it first, and after that do the switch to start using it.

[humitos]

This comment is valid for all the services we support.

[saadmk11]

Actually, I think we talked about adding the remote_id fields later, but now that I think of it it would be better to do it now, but I'll add it in another PR after this is merged.

[humitos]

Yeah. I said that in our conversation. However, I think it's better to have it before we do the initial resync so we already populate this data from the beginning and then we can start making use of it. I'm fine if it's in another PR.

[saadmk11]

This does not take into account for duplicate RemoteRepository for now.

Is this comment old now? The code seems to handle duplicated RemoteRepository properly now.

Unfortunately, it does not handle it. We can only filter RemoteRepository with the full name so that there is no duplication in the future, but the old/current duplication is still there so this will return multiple objects on get_or_create() which will result in an error

repo, _ = RemoteRepository.objects.get_or_create(
    full_name=fields['full_name']
)

We need to figure out a way to remove the current DB duplicates

[humitos]

We can only filter RemoteRepository with the full name so that there is no duplication in the future, but the old/current duplication is still there so this will return multiple objects on get_or_create() which will result in an error

I understand. We may not need to care about this anymore, if the what I'm going to describe next makes sense 😅

Talking to Eric last week, we thought that we can deploy all of these changes in a slightly different way than we originally described in #7536 (comment):

• between point 4 and 5 we could spin up a new instance with the new Python code and modelling (using RemoteRepository.Meta.db_table = "oauth_remoterepositoy_202011" so it creates a new table and doesn't touch the current one) and re-sync from VCS providers into that table. This new instance won't attend users requests, it will be used only for fetching data from VCS providers.
• after some days, once the re-sync is finished and we have all the data in the new table, we can dump the Project <-> RemoteRepository relations and load into the new table oauth_remoterepository202011. * finally, we do a full deploy for all our production instances with the new Python code and modelling

[saadmk11]

I understand. We may not need to care about this anymore, if the what I'm going to describe next makes sense

Talking to Eric last week, we thought that we can deploy all of these changes in a slightly different way than we originally described in #7536 (comment):

• between point 4 and 5 we could spin up a new instance with the new Python code and modelling (using RemoteRepository.Meta.db_table = "oauth_remoterepositoy_202011" so it creates a new table and doesn't touch the current one) and re-sync from VCS providers into that table. This new instance won't attend users requests, it will be used only for fetching data from VCS providers.
• after some days, once the re-sync is finished and we have all the data in the new table, we can dump the Project <-> RemoteRepository relations and load into the new table oauth_remoterepository202011. * finally, we do a full deploy for all our production instances with the new Python code and modelling

Sounds good 💯

[humitos]

I re-review this PR and everything looks fine to me. Thanks for your help on this! We can now move forward with the similar changes required for RemoteOrganization

[humitos]

This won't probably be required in the end as we may resync everything (pull down from VCS providers) a week in advance, but we can create a note and remove it later if needed.