Make ems_ref of service_port_config unique

[Ladas]

Make ems_ref of service_port_config unique by adding :protocol.

[miq-bot]

Checked commit Ladas@5eac3f6 with ruby 2.2.6, rubocop 0.47.1, and haml-lint 0.20.0
1 file checked, 0 offenses detected
Everything looks fine. 👍

[cben]

Sounds good, this is what we wanted to do, but I want to reproduce this in real life.
(Please do ping me if I neglect this!)

[simon3z]

@Ladas @cben why not just using the name? It's the unique identifier.

Also, I think that for correctness technically we'd need a migration (maybe we can get away without it, but at least let's consider it).

[Ladas]

@cben can we leverage the name here as @simon3z suggests?

[cben]

So just "#{service_id}_#{port_config.name}"? Sounds good.

@simon3z in ManageIQ/manageiq#15574 Ladislav concluded we're (sometimes) doing unnecessary deletes/inserts anyway so we can do them one last time, don't require a migration.
Port configs are not a user-visible collection, deletes/inserts should not cause problem.
By that logic, this is also backportable.

I keep thinking "why not first change refresh to ignore ems_ref and use individual parts", then drop ems_ref. Problem is that service_id string passed into parse_service_port_config() is not stored anywhere except in the ems_ref.
Wait, we don't need it, the port config is linked to specific parent service anyway!
Sounds like we can identify in old refersh just by [:name] and in graph refresh by :manager_ref => [:container_service, :name]?

[simon3z]

@simon3z in ManageIQ/manageiq#15574 Ladislav concluded we're (sometimes) doing unnecessary deletes/inserts anyway so we can do them one last time, don't require a migration.

👍

Sounds like we can identify in old refersh just by [:name]

Correct me if I'm wrong but it seems what I was suggesting in my previous comment, so 👍

[Ladas]

replaced by:
#168
ManageIQ/manageiq#16454