Instance is Not Removed from IG when node is marked as unschedulable

[freehan]

How to reproduce it (as minimally and precisely as possible):

kubectl cordon ${Node}

What happened:

Ingress controller did not remove the corresponding instance from IG and emit error messages like this:

E1220 14:43:16.052232  257911 instances.go:214] Failed to get zones for e2e-test-mixia-minion-group-3mzg: node not found e2e-test-mixia-minion-group-3mzg, skipping

What you expected to happen:
Unschedulable node got removed from instance group.

Root Cause:
Ingress controller failed to look up the corresponding zone of the instance when the node is marked as unschedulable. This is because the util function GetZoneForNode filter out the nodes that are not ready OR unschedulable. This is introduced here: #370

I think the GetZoneForNode should not filter out any nodes. It should return the zone of the node whenever is necessary. Thoughts?

/kind bug

[MrHohn]

#576 is attempting to remove the predicate filter.

[freehan]

closing this as #576 is merged