Tuning resources

In high throughput cluster, it is common practice to optimizing the allocation and usage of computing resources. In such environments, where the volume of transactions or operations per second is very high, resource tuning is essential to ensure that the cluster performs optimally.

The following is an example of how to use that practice in a large cluster. Further tweaks may required for different clusters.

agent:
  sensor:
    resources:
      requests:
        memory: 300Mi
        cpu: 160m
      limits:
        memory: 700Mi
        cpu: 800m

k8sWatcher:
  resources:
    requests:
      cpu: 50m
      memory: 300Mi
    limits:
      cpu: 100m
      memory: 500Mi

portal:
  resources:
    requests:
      cpu: 50m
      memory: 100Mi
    limits:
      cpu: 100m
      memory: 256Mi

clickhouse:
  resources:
    requests:
      cpu: 600m
      memory: 4096Mi
    limits:
      memory: 6000Mi

opentelemetry-collector:
  replicaCount: 2
  requests:
    cpu: 500m
    memory: 1024Mi
  limits:
    cpu: 1200m
    memory: 2048Mi

metrics-ingester:
  resources:
    limits:
      cpu: 750m
      memory: 512Mi
    requests:
      cpu: 250m
      memory: 256Mi

victoria-metrics-single:
  server:
    resources:
      requests:
        cpu: 1000m
        memory: 5000Mi
      limits:
        cpu: 1000m
        memory: 5000Mi

custom-metrics:
  extraArgs:
    remoteWrite.maxHourlySeries: "1000000"
    remoteWrite.maxDailySeries: "10000000"
  resources:
    requests:
      cpu: 500m
      memory: 512Mi
    limits:
      cpu: 1000m
      memory: 1Gi

Last updated