testcluster: Remove all dependencies on cluster spec protobuffer.
Instead, cluster nodepools are reconstructed from Kubernetes client API
data and looking at their labels from that. Then these are used for
configuring pods to schedule on these nodes.
This reduces dependencies on GKE-specific implementation details and
makes benchmarks easier to run outside of GKE.
testcluster
: Remove all dependencies on cluster spec protobuffer.Instead, cluster nodepools are reconstructed from Kubernetes client API data and looking at their labels from that. Then these are used for configuring pods to schedule on these nodes.
This reduces dependencies on GKE-specific implementation details and makes benchmarks easier to run outside of GKE.