ContainerServicePortConfig ems_ref not unique

[cben]

Found by @Ladas.
We set :ems_ref => "#{service_id}_#{port_config.port}_#{port_config.targetPort}".
Protocol is "TCP" / "UDP", and can have both on same ports. Here is an easy repro, we probably all have these:

> oc get service kubernetes -o json
...
    "metadata": {
        "uid": "fa61ef76-e6f7-11e6-a348-001a4a162683"
...
    "spec": {
        "ports": [
            {
                "name": "https",
                "port": 443,
                "protocol": "TCP",
                "targetPort": 443
            },
            {
                "name": "dns",
                "port": 53,
                "protocol": "UDP",
                "targetPort": 8053
            },
            {
                "name": "dns-tcp",
                "port": 53,
                "protocol": "TCP",
                "targetPort": 8053
            }
        ],
...

Can disambiguate by protocol.
Perhaps name would also work Nope, name is optional (https://kubernetes.io/docs/concepts/services-networking/service/; saw heapster service with unnamed port).

[cben]

@Ladas I do see 2 separate ContainerServicePortConfig records, with save_inventory and I think with graph refresh too.
Remind me, in which scenario do we expect this to cause a user-visible bug?

(save_container_port_configs_inventory uses only ems_ref as the find key.)

[Ladas]

@cben unless I am mistaken, the refresh would keep creating&deleting the duplicate ports, probably every other refresh would be create and other delete. But I would have to try to know for sure. :-) And it could randomly update the data with tcp or udp, since any of those would match.

[simon3z]

@cben @Ladas can you open a BZ? I feel we should resolve this before the graph refresh (in the current save_inventory).

[miq-bot]

This issue has been automatically marked as stale because it has not been updated for at least 6 months.

If you can still reproduce this issue on the current release or on master, please reply with all of the information you have about it in order to keep the issue open.

Thank you for all your contributions!

[cben]

long fixed on gaprindashvili and master.
graph refresh fixed long ago, too lazy to find now.
classic refresh fixed by ManageIQ/manageiq#16454
tests for all modes in ManageIQ/manageiq-providers-openshift#86