I am trying to figure out what is the trigger to scale AKS cluster out horizontally with nodes. I am having a cluster that runs on 103% CPU for 5+ minutes but there is no action taken. Any ideas what the triggers are and how I could customize them? If I start more jobs the cluster will lower the CPU allocation for all pods.
The article that MS has doesn't have anything specific around that https://learn.microsoft.com/en-us/azure/aks/cluster-autoscaler
You need to notice that:
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.
Which brings us to the actual Kubernetes 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 first condition above is the trigger you are looking for.
To get more details regarding the installation and configuration you can go through the Cluster Autoscaler on Azure. For example, you can customize your CA based on the Resources:
When scaling from an empty VM Scale Set (0 instances), Cluster
Autoscaler will evaluate the provided presources (cpu, memory,
ephemeral-storage) based on that VM Scale Set's backing instance type.
This can be overridden (for instance, to account for system reserved
resources) by specifying capacities with VMSS tags, formated as:
k8s.io_cluster-autoscaler_node-template_resources_<resource name>: <resource value>. For instance:
k8s.io_cluster-autoscaler_node-template_resources_cpu: 3800m
k8s.io_cluster-autoscaler_node-template_resources_memory: 11Gi
Related
I am running a service in AKS pods that would establish TLS connections with the client. There is a hard limit of 5K active connections per pod. I need a way to determine number of active TLS connections per pod and auto scale (HPA) when it reaches a threshold (say 3.5K TLS connections) and scale down when active connections are below 1K.
Is there a way to collect such metrics in AKS and scale based on that metrics. Kindly suggest.
By default, scale-up operations performed manually or by the cluster autoscaler require the allocation and provisioning of new nodes, and scale-down operations delete nodes. Scale-down Mode allows you to decide whether you would like to delete or deallocate the nodes in your Azure Kubernetes Service (AKS) cluster upon scaling down.
There is not any microsoft document that autoscale based on TLS
connection per pod.
Kubernetes has a cluster autoscaler, that adjusts the number of nodes based on the requested compute resources in the node pool. By default, the cluster autoscaler checks the Metrics API server every 10 seconds for any required changes in node count. If the cluster autoscale determines that a change is required, the number of nodes in your AKS cluster is increased or decreased accordingly. The cluster autoscaler works with Kubernetes RBAC-enabled AKS clusters that run Kubernetes 1.10.x or higher.
Cluster autoscaler is typically used alongside the horizontal pod autoscaler. When combined, 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.
To get started with the cluster autoscaler in AKS, see Cluster Autoscaler on AKS.
Reference : https://learn.microsoft.com/en-us/azure/aks/concepts-scale#cluster-autoscaler
For Counting the TLS connection to particuler nodes can be done using Plateform metrics ->Microsoft.Blockchain/blockchainMembers->ClusterCommEgressTlsConnectionCount
You can refer the same here.
In Azure K8s service, you can scale up the node pool but only we define the min and max nodes.
When i check the node pool scale set scale settings, i found it set to manual.
So i assume that the Node Pool auto scale does't rely on the belonging scale set, but i wonder, can we just rely on the scale set auto scale with the several metric roles instead of the very limited Node Pool scale settings ?
The AKS autoscaling works slightly different as the VMSS autoscaling.
From the official docs:
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 AKS autoscaler is tightly coupled with the control plane and the kube-scheduler, so it takes resource requests and limits into account that is far the better scaling method as the VMSS autoscaler (for k8s workload) that is anyway not supported for AKS:
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.
If an AKS cluster is created without zone-awareness (https://learn.microsoft.com/en-us/azure/aks/availability-zones#create-an-aks-cluster-across-availability-zones), what does this mean behind the scenes?
Are all the VMs running behind running in 1 of the 3 available availability zones?
When that zone has an outage, will Azure move the cluster to another AZ that is running?
This means that control plane components might (or will?) be in a single availability zone:
If you don't define any zones for the default agent pool when you
create an AKS cluster, the AKS control plane components for your
cluster will not use availability zones. You can add additional node
pools using the az aks nodepool add command and specify --zones for
those new nodes, however the control plane components remain without
availability zone awareness.
emphasis is mine. so this means even if your worker nodes will be up, your cluster won't work properly, because master nodes are down.
And no, it won't get moved to another availability zone in case of the outage
I have an AKS-Cluster in Azure. When I scale down the cluster with az aks scale for example I want to control which Node should be removed.
I cannot find a documentation that describes how Azure decides.
Will Azure prefer removing nodes that are already cordoned or drained?
Deleting it from the Azure Portal is not an option, because I want an application to communicate with Azure via CLI or API.
First of all, it's impossible to control which node to remove when you scale down the AKS cluster. Then I will show you how do the nodes change when you scale the AKS cluster.
When you do not use the VMSS as the agent pool, it means the AKS cluster use the individual VMs as the nodes. If you scale up, then it will increase the nodes with the index after the existing nodes. For example, the cluster has one node with the index 0 and then it will use the index 1 if you scale up one node. And if you scale down, it will remove the nodes with the biggest index in the sequence at first.
When you use the VMSS as the agent pool, it will comply with the scale rules of VMSS. And you can see the VMSS scale rules in the changes of VMSS scale up and down.
Also, you can take a look at the Azure CLI command az aks scale that scale the AKS cluster and the REST API.
I have create a VMSS in Azure Portal, to have the autoscale feature for my application. My application resided in Kubernetes cluster - around 10 microservices.
I want to create an Scale out rule, that if there is no enough memory , then increase the VM instance. But I don't see an option to set the rule based on memory. There are rules which we can define based on CPU utilization, disk space etc... But this won't help me to solve the problem. For my 10 microservice to work each service having 5 pods, i need to set a rule based on memory. If I set the rule based on CPU, the VM doesn;t scale up, as the CPU is not utilised much. Issue is with memory.
I get the error "0/3 nodes are available: 3 Insufficient pods.
The node was low on resource: [MemoryPressure]. "
I read that the memory rule is not available in host metrics in Azure, but it can enabled via guest metrics. To enable guest metrics, i see below link .
https://learn.microsoft.com/en-us/azure/virtual-machine-scale-sets/virtual-machine-scale-sets-mvss-guest-based-autoscale-linux
But I don't see an option to edit the template as defined in the above link. There is only "export Template" option visible for VMSS, where you cannot edit the template.
Could anyone please help me on this issue , to define memory rule for VMSS in Azure ?
No option seen to enable guest metrics for VMSS. No option to edit the template, only "export Template" option visible, where you cannot edit the template.
For the AKS autoscale, you just need to enable the autoscale function for your AKS cluster, set the min and max count of the nodes and then it will scale itself. You do not need to set the autoscale rule for it. Take a look at the AKS cluster autoscale.
When does Cluster Autoscaler change the size of a cluster?
Cluster Autoscaler increases the size of the cluster when:
there are pods that failed to schedule on any of the current nodes
due to insufficient resources.
adding a node similar to the nodes currently present in the cluster
would help.
Cluster Autoscaler decreases the size of the cluster when some nodes are consistently unneeded for a significant amount of time. A node is unneeded when it has low utilization and all of its important pods can be moved elsewhere.
And that what you have seen in the VMSS, the metric server is already installed in the high version AKS. If not install, you can install yourself and the steps here.