I’m running the latest Netdata in k8s with containerd runtime with cgroups enabled. The pod names aren’t resolving. All the names look like k8s system-slice containerd-service kubepods-best-effort pod- cri-containerd
I installed using vanilla helm chart.
Edit: I just discovered that it works with a node that has containerd 1.5.2 runtime but not with containers with containerd 1.4.6
Any advice on how to resolve?
Hi. Let’s check the following for both containerd 1.5.2 and 1.4.6
kubectl -n <NAMESPACE> exec -it <ND_CHILD_POD_NAME> -c netdata -- bash
apk add tree
# and share the output of this command
tree -d /host/sys/fs/cgroup
I’m on a closed network so I can’t move anything on my network to the internet. But when I hop onto each one - the 1.5.2 and the 1.4.6 I see they both have those directories available.
Also, 1.33 works but 1.35 doesn’t. The workaround for now is just use 1.33 on containerd 1.4.6 nodes.
I can’t really help if you don’t share
tree -d /host/sys/fs/cgroup (from inside the Netdata container) or
tree -d sys/fs/cgroup (from host).
We did some fixes regarding name resolution recently, but they haven’t been merged yet. I built a docker image that contains the changes, you can try it -
ilyam8/netdata-test-for-github. Let’s fee if it fixes the problem.
I appreciate your input but i have constraints on what I can do in the cluster and I can’t pull data off of it and I can’t run unapproved containers. For now i’ll just run 1.33 for runtime env containerd 1.4.6 and hope a) my cluster admins upgrade containerd to 1.5.2+ or b) netdata 1.36+ has a fix.