Is there a plan to add the ability to manually adjust the number of replicas on nginx controllers in app-routing-system?

[sohwaje]

I modified the "aks-app-routing-operator" deployment to adjust the number of replica, but it was impossible.

apiVersion: apps/v1
kind: Deployment
metadata:
  annotations:
    deployment.kubernetes.io/revision: "1"
  creationTimestamp: "2024-03-12T08:23:16Z"
  generation: 15
  labels:
    app.kubernetes.io/component: ingress-controller
    app.kubernetes.io/managed-by: aks-app-routing-operator
    app.kubernetes.io/name: nginx
  name: nginx
  namespace: app-routing-system
  ownerReferences:
  - apiVersion: approuting.kubernetes.azure.com/v1alpha1
    controller: true
    kind: NginxIngressController
    name: default
    uid: 6ea3e476-059e-44b3-9f8d-86655ee2a5dc
  resourceVersion: "113876214"
  uid: 37ea9332-baaf-47cf-aef4-f69d0ff2abef
spec:
  progressDeadlineSeconds: 600
  replicas: 2     <=============== I want to revise this

Currently, two nging controllers seem to be the default values.

kubectl get po -n app-routing-system
NAME                     READY   STATUS    RESTARTS   AGE
nginx-5d4cbcf56b-nrt46   1/1     Running   0          40h
nginx-5d4cbcf56b-swq58   1/1     Running   0          40h

Is there a plan to add the ability to manually adjust the number of replica on the nginx controller?

[MXClyde]

I have the seem need for a current project. In addition to being able to specify the resource requests on the pods (esp. CPU, which has a rather high default).

[OliverMKing]

Yes, you will be able to adjust the min/max replicas. You will also have a mechanism to adjust how aggressively the HPA scales. Expect this very soon.

I have the seem need for a current project. In addition to being able to specify the resource requests on the pods (esp. CPU, which has a rather high default).

Could you help us understand why you want to tweak the resource requests?

[MXClyde]

Yes, you will be able to adjust the min/max replicas. You will also have a mechanism to adjust how aggressively the HPA. Expect this very soon.

This is great!

Could you help us understand why you want to tweak the resource requests?

We are building an app platform on top of an autoscaling AKS cluster. In it's smallest -most cost-effective- incarnation we like to offer our customers the ability to run an app on a single-node 2-core cluster. Because the hardcoded nginx CPU request is 0.5 core, it is difficult to make our runtime fit in combination with 2 nginx replicas and the standard AKS containers. However on second thought, having the possibility of 1 nginx replica will likely suffice already. I can try that first when it is released.

[Duske]

Hey @OliverMKing, is there a roadmap when this feature will be rolled out?
We're currently want to set up a minimal environment for testing various pull requests, but when you choose a small VM with 2 vCPUs those 2 fixed replicas of the controller will claim 2 x 500m CPU for scheduling leaving only 1000m for all the other remaining pods. I didn't manage to override this setting, the operator always changes it back ♻️

[Duske]

Nevermind, just found out:

It's released you just need to upgrade to Kubernetes version 1.30 (or higher) to use it. Addon release policies mean we release new features behind new preview K8s versions first.

#244 (comment)

[polarn]

@Duske was the solution to just update the NginxIngressController CR? Or is it possible to provision this using the UI / az / terraform?

[Yarkane]

Hello there,
I understand that this is now possible to manually setup the number of replicas of app routing controllers.

However I agree with @MXClyde : 500m for each controller is huge, for example for a little development cluster. We would like to stay on app routing to keep the same environment as the production, and simply tweak the resources used by the controller.

@OliverMKing Is there anything in preview or in development to tweak the CPU and memory requests ?
An example in our dev cluster :

  Namespace                   Name                                                    CPU Requests  CPU Limits  Memory Requests  Memory Limits  Age
  ---------                   ----                                                    ------------  ----------  ---------------  -------------  ---
  app-routing-system          external-dns-787d4788dd-7dlqf                           100m (5%)     100m (5%)   250Mi (4%)       250Mi (4%)     2d10h
  app-routing-system          nginx-85497484cc-8fnvl                                  500m (26%)    100m (5%)   127Mi (2%)       60Mi (1%)      2d10h

[Duske]

@Duske was the solution to just update the NginxIngressController CR? Or is it possible to provision this using the UI / az / terraform?

Sorry, I missed your message. We did it by applying a patched CRD like this

apiVersion: approuting.kubernetes.azure.com/v1alpha1
kind: NginxIngressController
metadata:
  name: default
spec:
  controllerNamePrefix: nginx
  ingressClassName: webapprouting.kubernetes.azure.com
  scaling:
    maxReplicas: 1
    minReplicas: 1

As long as you can somehow apply these settings via UI, it should also work. With terraform this should definitely work as it's configuration-driven.