Recommendations for using Managed Service for Kubernetes

Use these recommendations for your PRODUCTION applications that require:

• High availability and fault tolerance.
• Load scaling.
• Resource isolation.

High availability and fault toleranceHigh availability and fault tolerance

• Use the REGULAR or STABLE release channel. NodeLocal DNS and Pod Topology Constraints are available in all Kubernetes versions.

• Update cluster and node groups manually. To do this, disable automatic master and node group updates.

• When creating a cluster, select the Regional master type. Kubernetes services will be available in the event of a failure at the availability zone level. [The Managed Service for Kubernetes Service Level Agreement] applies to the configuration with the regional master.

• Deploy multiple instances of services, such as Deployment and StatefulSet, in different availability zones. Use the Pod Topology Constraints and the AntiAffinity strategies to ensure high service availability and efficient usage of Kubernetes cluster resources.

  Use combinations of the labels below for all strategies:

  • topology.kubernetes.io/zone to keep the services available in case of availability zone failure.
  • kubernetes.io/hostname to keep the services available in the event of cluster node failure.

  Tip

  Test the strategies in a test environment before implementing them in PRODUCTION.

Load scalingLoad scaling

Use these recommendations if the load on your Managed Service for Kubernetes cluster is constantly growing:

• To make your cluster more robust, create node groups with automatic scaling in multiple availability zones.
• To reduce the load on the Kubernetes DNS, use NodeLocal DNS. If a cluster is made up of over 50 nodes, use automatic DNS scaling.
• To reduce horizontal traffic within a cluster, use the Network Load Balancer and the externalTrafficPolicy:Local rule if possible.
• Consider node storage requirements in advance:
  • Review disk limits for Yandex Compute Cloud.
  • Load test your disk subsystem in a test environment.
• To reduce latency at high IOPS, use non-replicated disks.

Isolating resourcesIsolating resources

Follow these recommendations for applications that use shared Kubernetes cluster resources.

Adjust the values of limits and requests for all cluster services:

---
...
containers:
...
  resources:
    limits:
      cpu: 250m
      memory: 128Mi
    requests:
      cpu: 100m
      memory: 64Mi
...

Specify vCPU availability in thousandths and RAM in megabytes. The service will not exceed the vCPU or RAM values specified in limits. Customized requests enable you to scale cluster nodes using automatic scaling.

To manage pod resources automatically, configure Kubernetes policies:

• Quality of Service for Pods to create pods in different availability classes.
• Limit Ranges to set limits at the namespace level.