Open draconiun opened 3 years ago
I had a go at this today. (Sorry it is a long time since you asked!)
GKE, at least when I tested it today, is running containerd, and the CRI socket can be accessed at /var/run/containerd/containerd.sock
. Therefore you need to change this argument on the agent:
- '--probe.cri.endpoint=unix:///var/run/containerd/containerd.sock'
You need to mount this path from the host into the container, like this:
volumeMounts:
- name: cri-socket
mountPath: /var/run/containerd/containerd.sock
volumes:
- name: cri-socket
hostPath:
path: /var/run/containerd/containerd.sock
type: Socket
Once I did that, Scope knew about my containers, however it thought they were all stopped. I can make a PR to fix that.
(Also that ds.yaml
file is out of date in other ways; it should use --probe.kubernetes.role=host
and run a separate probe for Kuebrnetes like in the k8s directory)
Hi, I currently have a kubernetes cluster deployed on GCP. The version is 1.19 and as an image type I use containerd. I have deployed weave scope version 1.13 and it does not show me the containers in the panel. And it is important to be able to visualize the CPU / Memory of my pods.
Pool in GKS:
I have checked this Weave Scope repository for CRI I tried to implement it but it still doesn't work. https://github.com/weaveworks/scope/blob/master/examples/cri/ds.yaml
I will be grateful for your responses. Thank you.