I am deploying a service to Azure Kubernetes Service.
The Horizontal Pod Autoscaler scales the number of pods, whereas the Cluster Autoscaler scales the number of nodes based on the number of pending pods. If my understanding is correct, if I don't set up autoscaling in my deployment file, the HPA won't get triggered, and only one pod will run; therefore, the CA won't get triggered either.
My question is - is there a scenario in AKS where the CA would get triggered, even without setting autoscaling in my deployment file?
My question is - is there a scenario in AKS where the CA would get triggered, even without setting autoscaling in my deployment file?
Cluster autoscaler is typically used together with the horizontal pod autoscaler. The Horizontal Pod Autoscaler increases or decreases the number of pods based on application demand, and the cluster autoscaler adjusts the number of nodes as needed to run those additional pods accordingly.
If your deployment does not have the capacity to automatically scale up or down via the HPA, NOR you don't manually increase number of pods to the level where no additional pods can run due to insufficient resource in your nodes then the CA would not be triggered therefore the answer is NO.
You might find this document from official azure docs helpful also.
Related
I am running a Kubernetes cluster in AWS EKS and I set up the autoscaler. I tested the autoscaler and it worked as when the number of pods in a node exceeded 110 then new nodes were automatically added to the cluster and the pending pods entered running state.
After that, I deleted the deployment. It's been about 10 minutes and I see that all new nodes created by the autoscaler are already there and in ready state!
How long does it take for Kubernetes to delete them automatically? Does it down-scale the cluster automatically at all?
Although scaling down is a slow process the default scan interval is 10 seconds if you are using the autoscaler to scale the nodes in EKS.
You can check the status of autoscaler using configmap and its a decision.
There could be a possibility that on the new node you have some system pod running so due to that EKS is not able to scale those nodes down or PDB(PodDisruptionBudget) is set for deployments.
Pod has the annotation "cluster-autoscaler.kubernetes.io/safe-to-evict": "false"
Read more about EKS scaling : https://docs.aws.amazon.com/eks/latest/userguide/autoscaling.html
when I change the replicas: x in my .yaml file I can see GKE autopilot boots pods up/down depending on the value, but what will happen if the load on my deployment gets too big. Will it then autoscale the number of pods and nodes to handle the traffic and then reduce back to the value specified in replicas when the request load is reduced again?
I'm basically asking how does autopilot horizontal autoscaling works?
and how do I get a minimum of 2 pod replicas that can horizontally autoscale in autopilot?
GKE autopilot by default will not scale the replicas count beyond what you specified. This is the default behavior of Kubernetes in general.
If you want automatic autoscaling you have to use Horizental Pod Autoscaler (HPA) which is supported in Autopilot
If you deploy HPA to scale up and down your workload, Autopilot will scale up and down the nodes automatically and that's transparent for you as the nodes are managed by Google.
GKE autoscale only Nodes by default, while you have to take care of your HPA deployment scaling.
Autopilot: GKE provisions and manages the cluster's underlying
infrastructure, including nodes and node pools, giving you an
optimized cluster with a hands-off experience.
We need to configure both scaling options for deployment VPA and HPA.
Pre-configured: Autopilot handles all the scaling and configuring of
your nodes.
Default: You configure Horizontal pod autoscaling (HPA) You configure
Vertical Pod autoscaling (VPA)
GKE will manage the scaling up/down of your nodes in node pools, without worrying about the infrastructure you just have to start deploying the application with HPA & VPA auto-scaling.
You can read more about the options here : https://cloud.google.com/kubernetes-engine/docs/concepts/autopilot-overview#comparison
In my 1 node AKS, I deploy multiple job resources (kind:jobs) that are terminated after the task is completed. I have enabled Cluster Autoscaler to add a second node when too many jobs are consuming the first node memory, however it scales out after a job/pod is unable to be created due to lack of memory.
In my job yaml I also defined the resource memory limit and request.
Is there a possibility to configure the Cluster Autoscaler to scale out proactively when it reaches a certain memory threshold (e.g., 70% of the node memory) not just when it cannot deploy a job/pod?
In Kubernetes you can find 3 Autoscaling Mechanisms: Horizontal Pod Autoscaler, Vertical Pod Autoscaler which both can be controlled by metrics usage and Cluster Autoscaler.
As per Cluster Autoscaler Documentation:
Cluster Autoscaler is a tool that automatically adjusts the size of the Kubernetes cluster when one of the following conditions is true:
there are pods that failed to run in the cluster due to insufficient resources.
there are nodes in the cluster that have been underutilized for an extended period of time and their pods can be placed on other existing nodes.
In AKS Cluster Autoscaler Documentation you can find note that CA is Kubernetes Component, not something AKS specific:
The cluster autoscaler is a Kubernetes component. Although the AKS cluster uses a virtual machine scale set for the nodes, don't manually enable or edit settings for scale set autoscale in the Azure portal or using the Azure CLI. Let the Kubernetes cluster autoscaler manage the required scale settings.
In Azure Documentation - About the cluster autoscaler you have information that AKS clusters can scale in one of two ways:
The cluster autoscaler watches for pods that can't be scheduled on nodes because of resource constraints. The cluster then automatically increases the number of nodes.
The horizontal pod autoscaler uses the Metrics Server in a Kubernetes cluster to monitor the resource demand of pods. If an application needs more resources, the number of pods is automatically increased to meet the demand.
On AKS you can adjust a bit your Autoscaler Profile to change some default values. More detail can be found in Using the autoscaler profile
I would suggest you to read the Understanding Kubernetes Cluster Autoscaling article which explains how CA works. Under Limitations part you have information:
The cluster autoscaler doesn’t take into account actual CPU/GPU/Memory usage, just resource requests and limits. Most teams overprovision at the pod level, so in practice we see aggressive upscaling and conservative downscaling.
Conclusion
Cluster Autoscaler doesn't consider actual resources usage. CA downscale or upscale might take a few minutes depending on cloud provider.
I need to create K8s autoscale setup for spark application which will be running - on premise and AWS both as docker images.By scale, I mean (scale up and down of nodes) from on-premise to AWS cloud using cluster autoscaler or by other means
I browsed so many articles like how to set up K8 cluster on AWS/ HPA & CA scaling but could not get concrete directions to follow
I am looking for any direction which can help me understand from where i should start/steps to follow to setup such K8s cluster.
Regarding Cluster Autoscaler:
Cluster Autoscaler is a tool that automatically adjusts the size of the Kubernetes cluster when one of the following conditions is true:
- there are pods that failed to run in the cluster due to insufficient resources,
- there are nodes in the cluster that have been underutilized for an extended period of time and their pods can be placed on other existing nodes.
The cluster autoscaler on Azure dynamically scales Kubernetes worker nodes. It runs as a deployment in your cluster.
This README will help you get cluster autoscaler running on your Azure Kubernetes cluster.
Regarding HPA:
The Horizontal Pod Autoscaler automatically scales the number of pods in a replication controller, deployment or replica set based on observed CPU utilization or other custom metrics. HPA normally fetches metrics from a series of aggregated APIs:
- metrics.k8s.io
- custom.metrics.k8s.io
- external.metrics.k8s.io
Metrics-server needs to be launched separately if you wish to base on something more than just CPU utilization. More info can be found here and here.
How to make it work?
HPA is being supported by kubectl by default:
kubectl create - creates a new autoscaler
kubectl get hpa - lists your autoscalers
kubectl describe hpa - gets a detailed description of autoscalers
kubectl delete - deletes an autoscaler
Example:
kubectl autoscale rs foo --min=2 --max=5 --cpu-percent=80 creates an autoscaler for replication set foo, with target CPU utilization set to 80% and the number of replicas between 2 and 5.
Here is a detailed documentation of how to use kubectl autoscale command.
I try to resize a kubernetes cluster to zero nodes using
gcloud container clusters resize $CLUSTER_NAME --size=0 --zone $ZONE
I get a success message but the size of the node-pool remains the same (I use only one node pool)
Is it possible to resize the cluster to zero?
Sometimes you just need to wait 10-20 minutes before autoscale operation takes effect.
In other cases, you may need to check if some conditions are met for downscaling the node.
According to autoscaler documentation:
Cluster autoscaler also measures the usage of each node against the node pool's total demand for capacity. If a node has had no new Pods scheduled on it for a set period of time, and all Pods running on that node can be scheduled onto other nodes in the pool, the autoscaler moves the Pods and deletes the node.
Note that cluster autoscaler works based on Pod resource requests, that is, how many resources your Pods have requested. Cluster autoscaler does not take into account the resources your Pods are actively using. Essentially, cluster autoscaler trusts that the Pod resource requests you've provided are accurate and schedules Pods on nodes based on that assumption.
Note: Beginning with Kubernetes version 1.7, you can specify a minimum size of zero for your node pool. This allows your node pool to scale down completely if the instances within aren't required to run your workloads. However, while a node pool can scale to a zero size, the overall cluster size does not scale down to zero nodes (as at least one node is always required to run system Pods)
Cluster autoscaler has following limitations:
- When scaling down, cluster autoscaler supports a graceful termination period for a Pod of up to 10 minutes. A Pod is always killed after a maximum of 10 minutes, even if the Pod is configured with a higher grace period.
Note: Every change you make to the cluster autoscaler causes the Kubernetes master to restart, which takes several minutes to complete.
However, there are cases mentioned in FAQ that can prevent CA from removing a node:
What types of pods can prevent CA from removing a node?
Pods with restrictive PodDisruptionBudget.
Kube-system pods that:
are not run on the node by default, *
don't have PDB or their PDB is too restrictive (since CA 0.6).
Pods that are not backed by a controller object (so not created by deployment, replica set, job, stateful set etc). *
Pods with local storage. *
Pods that cannot be moved elsewhere due to various constraints (lack of resources, non-matching node selectors or affinity, matching anti-affinity, etc)
*Unless the pod has the following annotation (supported in CA 1.0.3 or later):
"cluster-autoscaler.kubernetes.io/safe-to-evict": "true"
How can I scale my cluster to just 1 node?
Prior to version 0.6, Cluster Autoscaler was not touching nodes that were running important kube-system pods like DNS, Heapster, > Dashboard etc. If these pods landed on different nodes, CA could not scale the cluster down and the user could end up with a completely empty 3 node cluster. In 0.6, we added an option to tell CA that some system pods can be moved around. If the user configures a PodDisruptionBudget for the kube-system pod, then the default strategy of not touching the node running this pod is overridden with PDB settings. So, to enable kube-system pods migration, one should set minAvailable to 0 (or <= N if there are N+1 pod replicas.) See also I have a couple of nodes with low utilization, but they are not scaled down. Why?
How can I scale a node group to 0?
From CA 0.6 for GCE/GKE and CA 0.6.1 for AWS, it is possible to scale a node group to 0 (and obviously from 0), assuming that all scale-down conditions are met.
For AWS, if you are using nodeSelector, you need to tag the ASG with a node-template key "k8s.io/cluster-autoscaler/node-template/label/".
For example, for a node label of foo=bar, you would tag the ASG with:
{
"ResourceType": "auto-scaling-group",
"ResourceId": "foo.example.com",
"PropagateAtLaunch": true,
"Value": "bar",
"Key": "k8s.io/cluster-autoscaler/node-template/label/foo"
}