Question on scheduling with node affinity. “In” means I need a key and value, “ . . .

Kubernetes
#1

Bryan Tanoue:
Question on scheduling with node affinity. “In” means I need a key and value, “Exists” is only a key, “Does Not Exist” is anything not the key, “Not In” is Not the key and not the value. How does gt and lt work (is that greater then and less then?)

#2

Fernando Jimenez:
Let’s do run an example.

kubectl get node acgk8s-worker2 --show-labels | perl -pe 's/,/\n/g'
NAME             STATUS   ROLES    AGE    VERSION   LABELS
acgk8s-worker2   Ready    &lt;none&gt;   125m   v1.20.1   <http://beta.kubernetes.io/arch=amd64|beta.kubernetes.io/arch=amd64>
<http://beta.kubernetes.io/os=linux|beta.kubernetes.io/os=linux>
category=5
<http://kubernetes.io/arch=amd64|kubernetes.io/arch=amd64>
<http://kubernetes.io/hostname=acgk8s-worker2|kubernetes.io/hostname=acgk8s-worker2>
<http://kubernetes.io/os=linux|kubernetes.io/os=linux>

I labeled a node with the label catergory=5
Let’s run this manifest:

cat mypod.yaml
apiVersion: v1
kind: Pod
metadata:
  creationTimestamp: null
  labels:
    run: mypod
  name: mypod
spec:
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key: category
            operator: Gt
            values:
            - "3"
  containers:
  - image: nginx
    name: mypod
    resources: {}
  dnsPolicy: ClusterFirst
  restartPolicy: Always

Let’s check if the pod was spun in the node with the selected label.

kubectl get pod -o wide
NAME    READY   STATUS    RESTARTS   AGE    IP             NODE             NOMINATED NODE   READINESS GATES
mypod   1/1     Running   0          3m9s   192.168.19.4   acgk8s-worker2   &lt;none&gt;           &lt;none&gt;

Yes! It was!

#3

Bryan Tanoue:
That’s really cool!

#4

Bryan Tanoue:
I guess you can label a node with cpus or ram and make it easy to push out work loads assuming you know what is running where