hard20 pts · first answer countslinuxnetworkingkubernetes

Random packet drops on a busy node

Under peak load, a Kubernetes node randomly drops new connections. CPU and memory are fine. dmesg shows the line below. What is the right fix?

nf_conntrack: nf_conntrack: table full, dropping packet

A.Increase the pod memory limitB.Restart kube-proxy every hourC.Turn off the node firewall completelyD.Raise nf_conntrack_max (sized to memory), shorten idle timeouts, and reduce short-lived connections with keep-alive

← Alerting on queue lag Listen queue overflows →