tencent cloud

High Availability Configuration Optimization
Download PDF
Last updated: 2024-08-12 17:48:23
High Availability Configuration Optimization
Last updated: 2024-08-12 17:48:23
Download PDF

Overview

This document describes the high-availability deployment configuration methods for Nginx Ingress.

Increasing the Number of Replicas

Configure automatic scaling:
controller:
  autoscaling:
    enabled: true
    minReplicas: 10
    maxReplicas: 100
    targetCPUUtilizationPercentage: 50
    targetMemoryUtilizationPercentage: 50
    behavior: # Quick scale-out to handle traffic peaks, slow scale-in to leave a buffer to avoid abnormal traffic
      scaleUp:
        stabilizationWindowSeconds: 300
        policies:
          - type: Percent
            value: 900
            periodSeconds: 15 # Allowing scale-out up to 9 times the current number of replicas every 15 seconds
      scaleDown:
        stabilizationWindowSeconds: 300
        policies:
          - type: Pods
            value: 1
            periodSeconds: 600 # Allowing scale-in of only one pod at most every 10 minutes
If you want a fixed number of replicas, directly configure replicaCount:
controller:
  replicaCount: 50

Spreading Scheduling

Use topology distribution constraints to spread out pods for disaster recovery and avoid single points of failure:
controller:
  topologySpreadConstraints: # Policy to maximize spreading
    - labelSelector:
        matchLabels:
          app.kubernetes.io/name: '{{ include "ingress-nginx.name" . }}'
          app.kubernetes.io/instance: '{{ .Release.Name }}'
          app.kubernetes.io/component: controller
      topologyKey: topology.kubernetes.io/zone
      maxSkew: 1
      whenUnsatisfiable: ScheduleAnyway
    - labelSelector:
        matchLabels:
          app.kubernetes.io/name: '{{ include "ingress-nginx.name" . }}'
          app.kubernetes.io/instance: '{{ .Release.Name }}'
          app.kubernetes.io/component: controller
      topologyKey: kubernetes.io/hostname
      maxSkew: 1
      whenUnsatisfiable: ScheduleAnyway

Scheduling to Dedicated Nodes

Typically, the load of Nginx Ingress Controller is proportional to the traffic. Considering its importance as a gateway, we recommend scheduling it to dedicated nodes or super nodes to avoid interfering with business pods or being interfered by them.
Schedule it to the specified node pool:
controller:
  nodeSelector:
    tke.cloud.tencent.com/nodepool-id: np-********
Note:
Super nodes perform better as all pods exclusively occupy the virtual machine without mutual interference. If you are using a serverless cluster, there is no need to configure the scheduling policy here, as it will only be scheduled to super nodes.

Setting Reasonable requests and limits

If Nginx Ingress is not scheduled to super nodes, set requests and limits reasonably to ensure sufficient resources while avoiding excessive resource usage that leads to high node load:
controller:
  resources:
    requests:
      cpu: 500m
      memory: 512Mi
    limits:
      cpu: 1000m
      memory: 1Gi
If you are using super nodes or a serverless cluster, you need to define requests only, that is, declare the virtual machine specifications for each pod:
controller:
  resources:
    requests:
      cpu: 1000m
      memory: 2Gi

Was this page helpful?
You can also Contact Sales or Submit a Ticket for help.
Yes
No

Feedback