k8s.md

kubectl --context=staging get nodes

kubectl --context=staging get pods

Show secrets:

kubectl get secrets --all-namespaces kubectl get secret the-secret-name -o jsonpath='{.data}' | jq kubectl get secret the-secret-name -o jsonpath='{.data.foo}' | base64 --decode

Get pods matching a label:

kubectl get pods -l app=foo

Useful image for debugging problems:

k debug mypod -it --image=nicolaka/netshoot

For k3s

Lots of details:

sudo k3s check-config

To stop all the containers (since sudo systemctl stop k3s doesn't do it, per k3s-io/k3s#2583):

sudo /usr/local/bin/k3s-killall.sh

This is good if you're trying to give the ports back to an nginx running on the host.

To move pods from one node group to another:

The goal is to launch pods on the new nodes before killing the pods on the old nodes.

• Create the new node group. Wait until the nodes are ready.
• Taint the old nodes so no new pods will get scheduled there:

  k taint nodes ip-10-10-245-95.ec2.internal key1=value1:NoSchedule

• Scale the deployment to launch new pods, which should all get scheduled on the new nodes:

k scale deployment --replicas=2 -n foo bar

• Drain the old nodes:

k drain ip-10-10-245-95.ec2.internal --ignore-daemonsets --delete-emptydir-data

• Scale down:

k scale deployment --replicas=1 -n foo bar