Container startup gets stuck in vnc_startup.sh when running on AWS ECS, on EC2 instances (not with Fargate) and with awsvpc network mode

[fj604]

workspaces-core-images/src/common/startup_scripts/vnc_startup.sh

Lines 466 to 493 in ff3eb72

	function wait_for_network_devices() {
	while true; do
	interfaces=$(ip link show type veth | awk -F: '/^[0-9]+: / {print $2}' | awk '{print $1}' | sed 's/@.*//')
	if [ -z "$interfaces" ]; then
	sleep 1
	continue
	fi
	for interface in $interfaces; do
	# ignore eth* interfaces if egress gateway is enabled
	if [[ $interface == eth* && -z $KASM_SVC_EGRESS ]]; then
	return
	fi
	if [[ $interface == k-p-* ]]; then
	wait_for_egress_signal
	if [ -z "$KASM_PROFILE_LDR" ]; then
	http_proxy="" https_proxy="" curl -k "https://${KASM_API_HOST}:${KASM_API_PORT}/api/set_kasm_session_status?token=${KASM_API_JWT}" -H 'Content-Type: application/json' -d '{"status": "running"}'
	fi
	return
	fi
	done
	sleep 1
	done
	}

When running KASM workspaces in AWS ECS on EC2 (not Fargate) and using awsvpc network mode, the container gets stuck in startup script and never starts listening on port 6901.

default:~$ ip link show type veth
3: ecs-eth0@if8: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default
    link/ether 0a:58:a9:fe:ac:03 brd ff:ff:ff:ff:ff:ff link-netnsid 0

• The script expects interfaces be of type venv and to be named eth*
• In awsvpc mode the container is given its own elastic network interface (ENI) that appears as ecs-eth0.
• Because ecs-eth0 does not start with eth nor with k-p-, neither branch of the test fires, the loop finishes one iteration, sleeps one second, and repeats indefinitely. Start-up therefore stalls even though the interface was correctly “seen”.