Thanks — I can help troubleshoot this. You mentioned voice stops working after changing the assigned private IP address on your EKS nodes. To narrow down the cause, I need a bit more detail about your networking and LiveKit configuration so I can give targeted steps. Could you share the following? 1. Networking setup ◦ Are you using an NLB/ALB, Ingress, or direct nodePort to expose LiveKit? ◦ Are your nodes in a private subnet with NAT, or do they have public IPs? ◦ Did you change the node primary IPs, assign additional ENIs, or modify pod networking (CNI) IPs? 1. LiveKit & cluster config ◦ Which LiveKit components are running in the cluster (server only, server + web, TURN relay)? ◦ How is LiveKit configured for advertised IPs / interfaces? Do you set

public_ip

/

advertised_ip

or any related env vars in LiveKit config? 1. NAT / STUN / TURN details ◦ Are you using TURN (recommended for NAT traversal) and is it reachable after the IP change? ◦ Which STUN/TURN server are you using (embedded, external), and does it bind to a specific IP? 1. Symptoms & logs ◦ When you try, do clients connect but audio is missing, or do connections fail entirely? ◦ Any error messages in LiveKit server logs, browser console (getUserMedia/ICE/DTLS/SRTP), or client SDK logs? ◦ Any network capture or

kubectl logs

lines you can paste (especially ICE negotiation or TURN allocation errors)? 1. Timing and change details ◦ Exactly what change did you make and how (AWS console, terraform, CNI reconfiguration)? ◦ Did you redeploy pods after the IP change or restart LiveKit components? What you can try right away • Verify STUN/TURN reachability from a client and from within a pod (curl/ntp not useful for TURN — try allocating a relay or test with

trickle-ice

pointing to your STUN/TURN). • Check LiveKit server logs for ICE/TURN errors, e.g., failure to allocate relay or invalid ICE candidates. • Ensure

advertised_ip

(or equivalent) matches the IP clients should reach: if node IPs changed, LiveKit may still advertise the old IP to clients, causing connectivity failure. • If you use a load balancer, confirm its target group/backends reflect the new node IPs and health checks succeed. If you share the above details (especially relevant log lines and how LiveKit advertises its address), I’ll propose exact configuration fixes and commands to run to restore voice. , with interactive elements