OpenShift nodes not evicting pods that use too much memory

Issue

• Per the soft eviction setting, eviction should have taken place to reclaim memory but we don't see any eviction here. The node reports insufficient memory and the following error is repeated many times in the logs:

  Feb 02 13:57:49 node.example.com atomic-openshift-node[73341]: I0202 13:57:49.948212   73341 helpers.go:800] eviction manager: eviction criteria not yet met for threshold(signal=memory.available, operator=LessThan, value={2Gi 0}, gracePeriod=1m30s), duration: 10.037656746s