How can I monitor performance and resource usage on my Kubernetes (LKE) cluster?

There are a few ways you can gather insight for your cluser's performance with LKE.

LKE

For starters, you can run these two commands right out-of-the-box with LKE. They provide valuable insight the the status of your cluster.

$ kubectl describe pods
$ kubectl describe nodes

However, it seems that you may be looking for more real-time insight, so I've included some more options below.

Metrics-Server

I can't say that metrics-server will come pre-installed with LKE. However, you can certainly install this on your own and use kubectl top to view resource metrics from your cluster. Here are some resources that may be helpful in setting this up:

Bitnami: Metrics-Server

GitHub: Metrics-Server

If you follow the second guide, you'll need to upgrade metrics-server since metrics.k8s isn't available in our current cluster. After installation, the following command should get you on your way:

$ helm upgrade my-release bitnami/metrics-server \
    --set apiService.create=true

Prometheus

Another option you may be interested in exploring is Prometheus. This tool collects time series data, then allows you to visualize metrics and configure alerts. This guide has instructions that you may follow to get this set up: GitHub: Prometheus

The following command should get you started:

$ helm install [name] stable/prometheus

Grafana

Here's another similar tool for monitoring your cluster. Grafana offers data visualization which allows you to generate graphs and maps to understand your data. You may also set up alerts to be notified when certain conditions are met. Here's the page for GitHub: Grafana. To kick things off, you can run the following command to start the installation process:

$ helm install [name] grafana stable/grafana

Hope this helps!

Grafana requires RWX storage, though, which the Linode CSI driver currently doesn't support, AFAICT?

That's correct @catalina — at the moment, the Linode CSI driver doesn't support directly ReadWriteMany (RWX) access for persistent volumes.

However, Grafana only requires RWX access for scaling. On a single pod, it should work fine.

Otherwise, if you do wish to scale your Grafana deployment, you can create a RWX-accessible Block Storage Volume using the NFS Server Provisioner helm chart. For example, this command would create a 100GiB Volume for the StorageClass nfs that could support RWX-accessible PVC's:

helm install nfs-server stable/nfs-server-provisioner \
--set persistence.enabled=true,persistence.storageClass=linode-block-storage-retain,persistence.size=100Gi

You could then set up a PVC with RWX with the following manifest:

---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: nfs-pvc
spec:
  accessModes:
  - ReadWriteMany
  resources:
    requests:
      storage: 10Gi
  storageClassName: nfs

Setting claimName: nfs-pvc for in your deployment's PVC spec would then enable you to scale your deployment while utilizing the same persistent data.

Quick follow up! Here's another post that details installing metrics server on LKE:

How Can I Deploy the Kubernetes-Metrics Server on LKE?