Eviction of cattle-node-agent

I have cluster with 50+ nodes and eviction policy set to:
eviction-hard: memory.available<1.6Gi.

But first thing to be evicted is always cattle-node-agent or cluster-agent etc. What is recommended way to deal with this? Manually patching memory request for these deamonsets, setting priority class?

  Type     Reason     Age   From               Message
  ----     ------     ----  ----               -------
  Normal   Scheduled  15m   default-scheduler  Successfully assigned cattle-system/cattle-node-agent-h8tpv to worker15
  Normal   Pulled     15m   kubelet            Container image "harbor/rancher/rancher-agent:v2.6.0" already present on machine
  Normal   Created    15m   kubelet            Created container agent
  Normal   Started    15m   kubelet            Started container agent
  Warning  Evicted    14m   kubelet            The node was low on resource: memory. Container agent was using 48248Ki, which exceeds its request of 0.
  Normal   Killing    14m   kubelet            Stopping container agent