[sig-node] node-lifecycle detects unexpected not ready node
[sig-node] node-lifecycle detects unreachable state on node

Two above tests fail on azure upgrade job around 20% on avg due to networking issue. According to initial investigation, established TCP connections between Kubelets and the Azure balancer VIP are black-holing. This issue may be related to hairpin NAT issue with Azure LB even though it was supposed to be addressed via the custom routes script on master in machine-config-operator.

    4.18

    Run rollout or upgrade job for Azure on OCP 4.18

    1.
    2.
    3.
    

Here is the doc on the networking issue: https://docs.google.com/document/d/1rYjqiEs3hlOZwL_L5oKy_WcLvAzZoi_pu3023QqU4CE/