Working with Node Selector & Node Affinity In Kubernetes

Node Selector & Node Affinity in Kubernetes

@pramodAIML
Oct 23, 2020 · 5 min read
Image for post
Image for post

“If you stretch beyond your capacity without putting your wisdom into practice, you are bound to break ”

The same applies to our Kubernetes worker nodes. If you will try to schedule a heavy workload to the worker node of low capacity, there is a higher probability of this node filling fast and create inefficiency in the cluster. So as a Kubernetes administrator or developer what you should do?

To understand this scenario let me paint a visual picture of a hypothetical situation below:

Explanation:

In fig 1.0, there are three worker node, NODE 1, NODE 2, NODE 3.

  • NODE 1 is a high-end machine which has high storage and memory capacity
  • NODE 2 is a medium-end machine with a medium level of storage and memory capacity.
  • NODE 3 is a low-end machine with a low level of storage and memory capacity.

Also, there are three pods with

  • POD 1 with a Very heavy workload
  • POD 2with a Medium level workload
  • POD 3 with the lowest workload

Now by default Kubernetes may end up in the situation depicted below in fig 2.0

Explanation:

As shown in fig 2.0, there can be a situation that POD 1 with the heaviest workload may end up being scheduled in NODE 3 which has the lowest capacity and smaller workload pod may end up being scheduled in NODE 1 or NODE 2 for that matter. Which you as an administrator may not want to happen.

So to handle this kind of uncertainties there is a concept of

  • Node Selector
  • Node Affinity

Let’s get into the details of both solutions

What Is Node Selector & How Does It Work?

As explained above, you may want to schedule a particular type of pod with a particular type of workload to be scheduled only to a particular type of Node selectively.

Generally, such constraints are not required, as our scheduler is intelligent enough to automatically do a reasonable placement to avoid placing the pod on a node with insufficient free resources, but there are some instances where you may want more control on a node where a pod lands, for example

  • You may want to ensure that any given pod should end up on a machine with an SSD attached to it
  • Or You may desire to co-locate pods from two different services that communicate a lot into the same availability zone.

The node selector is one such mechanism or constraint which we can apply to our pod to ensure that it is placed into a particular type of Node. One has to specify the field

nodeSelector”,

which comes under the Pod’s Spec definition. This field makes use of key-value pairs. Any pod to be eligible for running onto any specific Node has to have the specified key-value pairs as labels of the intended Node.

How to apply nodeSelector?

Step1: First Label the Node :

Get the name of the node

To label, the node, use the following command to get all the nodes you have in your K8S cluster

$ kubectl get nodes

So if you run the above command you will get to see at least one node which will be your master node, as shown below.

Image for post
Image for post

Here is my case: the highlighted area depicts our master node. we can look into the details of this node using

$ kubectl describe nodes

You can see that this minikube is our default master node has some pre-defined labels.

Attach Labels To The Node You Want :

Here we only have master node minikube so we will label it, using the command below:

$ kubectl label node minikube size=large

Once you apply this command, a node with node name (minikube )will be labeled with key-value pairs:

“ size=large “

Let’s view the Node label, To See if our label has been mapped successfully, by typing the command:

$ kubectl describe nodes

output:

If you carefully observe the above output, the highlighted area shows that our node minikube has been labeled with key-value pair, size=large

Now that we have labeled our Node, it’s time to create a Pod which will be making use of nodeSelector field to define this key-value pair

Step2: Create a Pod definition and make use of nodeSelector ,

as shown below in the :node-sel-demo.yaml:

apiVersion: v1
kind: Pod
metadata:
name: nginx
labels:
env: staging
spec:
containers:
- name: nginx
image: nginx

nodeSelector:
size: large

Step3: Create this Pod using the apply command as shown below

$ kubectl apply -f nod-sel-demo.yaml

this will successfully create the pod which has been scheduled to run on our minikube master node based on the nodeSelector , key-value parameter as shown below

nodeSelector:
size: large

Let’s verify if our pod has been scheduled on our master node or not, using the below command:

$ kubectl get pods -o wide

output:

it can be clearly seen by the highlighted area that our pod nginx, has been scheduled on master node minikube.

Node Selector Limitation:

Now that we have learned how to label any specific node and then use that label to bind any particular Pod to be specifically scheduled on this node using nodeSelector(key-value pair),

it is imperative to understand that if you want to run our Pod on a particular node based on the following conditions, like

  • size: large or medium
  • size: Not Small

This type of logical expression type of selection cannot be achieved by nodeSelector, for this one has to use

“ Node Affinity “

We will look into the details of nodeAffinity,

in our next section where we will cover

  • What Is Node Affinity?
  • What is Affinity?
  • How Node Affinity Works?
  • Example Pod definition to understand nodeAffinity application

Till then, keep reading, keep supporting, and if you all are loving my contribution, don’t forget to follow, clap, and share.

Bye. Bye.. and Take care…….

SysopsMicro

Learn System Designing | Architecting | DevOps| Microservices | Clean Coding | Data Science

@pramodAIML

Written by

Everything About Artificial Intelligence | Machine Learning | Applied statistics| Data Science. Let’s learn AI & ML together

SysopsMicro

If you want to learn how to code and maintain scalable tech architecture, you have come to the right place. We are a team of consultant who spwcialise in clean coding, programming, system designing & archtecting, Decomposing Monolith to Microservices and Impelmenting Ci-CD

@pramodAIML

Written by

Everything About Artificial Intelligence | Machine Learning | Applied statistics| Data Science. Let’s learn AI & ML together

SysopsMicro

If you want to learn how to code and maintain scalable tech architecture, you have come to the right place. We are a team of consultant who spwcialise in clean coding, programming, system designing & archtecting, Decomposing Monolith to Microservices and Impelmenting Ci-CD

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store