Check reconciliation of routes

[jpkrohling]

We need to check whether anything changed related to the reconciliation of routes. We don't seem to be getting an existing route during the reconciliation of the Jaeger CR, which causes the loop to fail while attempting to create a new route:

time="2020-02-04T15:40:03Z" level=debug msg="Reconciling Jaeger" execution="2020-02-04 15:40:03.66812891 +0000 UTC" instance=with-cassandra namespace=fred
time="2020-02-04T15:40:29Z" level=debug msg="creating route" instance=with-cassandra namespace=fred route=with-cassandra
time="2020-02-04T15:40:29Z" level=debug msg="Reconciling Jaeger completed" execution="2020-02-04 15:40:03.66812891 +0000 UTC" instance=with-cassandra namespace=fred

time="2020-02-04T15:40:29Z" level=debug msg="Reconciling Jaeger" execution="2020-02-04 15:40:29.867572629 +0000 UTC" instance=with-cassandra namespace=fred
time="2020-02-04T15:40:29Z" level=debug msg="creating route" instance=with-cassandra namespace=fred route=with-cassandra
time="2020-02-04T15:40:30Z" level=error msg="failed to apply the changes" error="routes.route.openshift.io \"with-cassandra\" already exists" execution="2020-02-04 15:40:29.867572629 +0000 UTC" instance=with-cassandra namespace=fred

cc @kevinearls, @rubenvp8510

[jpkrohling]

In the end, this might be just a timing issue, as further reconciliations on the same CR do work:

time="2020-02-04T15:40:31Z" level=debug msg="Reconciling Jaeger" execution="2020-02-04 15:40:31.01279533 +0000 UTC" instance=with-cassandra namespace=fred
time="2020-02-04T15:40:44Z" level=debug msg="updating route" instance=with-cassandra namespace=fred route=with-cassandra
time="2020-02-04T15:40:44Z" level=debug msg="Reconciling Jaeger completed" execution="2020-02-04 15:40:31.01279533 +0000 UTC" instance=with-cassandra namespace=fred

As Kubernetes is "eventually consistent", we should consider that the route object isn't being created as fast as other objects, so, it doesn't indeed exist during the second reconciliation.

[jpkrohling]

I think this has been fixed (or mitigated) by using non-cached clients, from #937.