Matthew Robinson:
What network CNI layer are you using?
Timekiller:
Oh hey Matthew! Fancy meeting you here
The same one being used in the demo: https://www.weave.works/docs/net/latest/kubernetes/kube-addon/
Timekiller:
Does weave net expect the docker container engine? I just noticed this in the weave logs:
INFO: 2022/06/01 05:22:22.457418 Command line options: map[conn-limit:200 datapath:datapath db-prefix:/weavedb/weave-net docker-api: expect-npc:true http-addr:127.0.0.1:6784 ipalloc-init:consensus=0 ipalloc-range:10.32.0.0/12 metrics-addr:0.0.0.0:6782 name:2e:59:61:6c:f0:bd nickname:ip-172-31-28-82.ec2.internal no-dns:true no-masq-local:true port:6783]
I’m using containerd
Matthew Robinson:
I just had a bit of a Google and it looks like a few people are having issues with containerd and WeaveNet
Matthew Robinson:
https://serverfault.com/questions/1100808/pod-coredns-stuck-in-containercreating-state-with-weave-on-k8s
Timekiller:
Yeah, I’ve seen that first one.
Timekiller:
Any alternative CNI you’d recommend? There are a bunch
Timekiller:
I’m using the same version of containerd as well 1.6.4
Matthew Robinson:
When I started playing I used flannel as it is probably the simplest of the CNIs
Timekiller:
That’s an excellent suggestion. Thank you so much! I’ll do that next
Matthew Robinson:
If I remember correctly flannel can’t do some things, network policies rings a bell
Timekiller:
That’s good to keep in mind! I can always switch later if flannel works
Matthew Robinson:
I’m mostly using EKS at the moment with the AWS VPC CNI which I wouldn’t recommend for experimenting
Timekiller:
No worries, I’ll stay away from EKS! Still planning to move to a more managed solution later on after I’m done playing with kubeadm
setup
unnivkn:
Hi @Timekiller the root cause here is while you install kubeadm you might have passed --pod-network-cidr= <some IP Rang>
. So this Ip range should match with the default IP range of the CNI you are trying to install. For eg: weavenetwork use 10.x.x.x Ip range & Flannel uses 172.x.x.x IP range. So if you use Flannel Ip range(eg: 172.1.0.22) along with kubadm command & try to install weave network CNI you may get the error, which you already encountered above. fyr:
Timekiller:
Hello @unnivkn - I’m not really sure that’s the root cause.
Did you see the <Issues · containerd/containerd · GitHub issue> linked earlier in this thread? That’s a very clear description of the problem I’m experiencing which seems to be an issue with containerd
version 1.6.4
& weave
A new version of containerd
just came out 1.6.5
and I decided to try spinning it up with weave to see if there is any change, and… sure enough, the issue seems to be fixed!
Have a look:
Timekiller:
These two EC2 instances are identical, with the following userdata script ran on both of them (weave/flannel installation steps are commented out and performed manually)