https://rancher.com/ logo
Title
c

crooked-rocket-3338

07/19/2022, 2:51 PM
Hi guys, is the issue related to kube-router not cleaning up IP (https://github.com/rancher/rancher/issues/14504) is still happening ? I have a k3s cluster (1.22) and I realized some of my load balancer failed with the
failed to allocate for range
thing.
Do you guys know a cleaner way to handle this rather that
rm -rf /var/lib/cni/networks/cbr0/*
then reboot ?
c

creamy-pencil-82913

07/19/2022, 4:44 PM
That issue is with RKE. Are you using RKE, or K3s?
K3s doesn’t use kube-router for anything but network policy, it is not involved in the CNI stack at all.
c

crooked-rocket-3338

07/19/2022, 6:46 PM
I am using k3s so this is definitely not the same issue. Do you know what can lead to
failed to allocate for range 0: no IP addresses available in range set: 10.42.1.1-10.42.1.254
error ?
c

creamy-pencil-82913

07/19/2022, 7:21 PM
How many pods do you have running on the node?
did you raise MaxPods without changing the cidr size?
c

crooked-rocket-3338

07/19/2022, 7:22 PM
nope I got 20/25 pods but still not IP available after a while.
c

creamy-pencil-82913

07/19/2022, 7:22 PM
I’m confused as to why it would even be allocating pod ranges for a load balancer. ServiceLB doesn’t allocate IPs, it just uses the node IPs. Did you deploy your own LB controller?
It might be helpful if you opened a GH issue in the k3s repo and attached the complete service logs from the node in question
sometimes if you just google a snippet of an error message you can end up troubleshooting a completely unrelated issue
👀 1
c

crooked-rocket-3338

07/19/2022, 7:24 PM
pretty much everything is stock default k3s unless I did something completely wrong without noticing. But I try not to bring anything that "is my own" 🙂
Ok will try to gather more info and open a GH issue.