k3s-host-network-experiment
Why did my minidlna setup stop working?
Update 1.
securityContext:
capabilities:
add:
- NET_ADMIN
Enables SSDP/udp multicast on :1900.
Commands
bin/k3s/uninstall
bin/k3s/<cni>/<strategy>/install
Matrix
cni | config | notes |
---|---|---|
flannel | default | |
flannel | custom | INSTALL_K3S_EXEC="--disable-network-policy --no-deploy traefik --bind-address=$(bin/ip) --node-ip=$(bin/ip) --node-external-ip=$(bin/ip) --cluster-cidr=192.168.0.0/16 --service-cidr=192.168.0.0/16" |
none | custom | INSTALL_K3S_EXEC="--flannel-backend=none --no-flannel --disable-network-policy --no-deploy traefik" |
cilium | default | INSTALL_K3S_EXEC="--flannel-backend=none --no-flannel --disable-network-policy --no-deploy traefik" |
Debug
ss -lntp | grep -E ':(8200) '
netstat -lntp
Capture
tshark -f "udp port 1900"
Discovery
nmap -sU -p 1900 --script=upnp-info 10.0.0.0/24
Notes
-
Yes, you cannot access pods from outside the cluster when using flannel - they're on virtual interfaces with private addresses. This is normal. k3s-io/k3s#1879 (comment)
- udp/upnp multicast (not broadcast) on 1900.
- tcp/http serve on 8200.
- can generally access internally from host ip e.g. http://10.0.0.238:8200/
- external port block on :8200? iptables, ufw, ...
- k3s installer options