Skip to content

Etcd configuration#

k0smotron in HCP mode deploys an etcd cluster to store the state of the control plane. The etcd cluster is deployed as a StatefulSet with odd number of replicas to ensure quorum. For example:

Control Plane replicas etcd replicas
1 1
2 3
3 3
4 5
5 5
… …

Default configuration#

The etcd cluster is deployed with the following default configuration options:

etcd flag default value
--auto-compaction-mode=periodic periodic
--auto-compaction-retention 5m
--snapshot-count 10000

Customizing etcd configuration#

k0smotron supports passing (and overriding default) etcd configuration flags to the etcd cluster by setting the spec.etcd.args field in the Cluster resource. For example:

apiVersion: k0smotron.io/v1beta1
kind: Cluster
metadata:
  name: k0smotron-test
spec:
  replicas: 3
  etcd:
    args:
      - --auto-compaction-mode=periodic
      - --auto-compaction-retention=15m
      - --snapshot-count=20000

Defragmentation#

k0smotron supports running etcd defragmentation job periodically. The defragmentation job is disabled by default. To enable the defragmentation job, set the spec.etcd.defragJob.enabled field to true in the Cluster resource. If enabled, by default the defragmentation job runs every day at 12:00. You can customize the schedule and the rule for the defragmentation job by setting the spec.etcd.defragJob.schedule and spec.etcd.defragJob.rule fields in the Cluster resource. For example:

The feature is based on the etcd-defrag tool. For more information, see the etcd-defrag repo.

apiVersion: k0smotron.io/v1beta1
kind: Cluster
metadata:
  name: k0smotron-test
spec:
  etcd:
    defragJob:
      enabled: true
      schedule: "0 0 * * *" # Default: 0 12 * * *
      rule: "dbQuotaUsage > 0.5 || dbSize - dbSizeInUse > 200*1024*1024" # Default: dbQuotaUsage > 0.8 || dbSize - dbSizeInUse > 200*1024*1024

Resource Requirements#

k0smotron supports setting resource requirements (requests and limits) for the etcd StatefulSet pods. By default, etcd pods are created with no specific resource requirements. To set resource requirements, use the spec.etcd.resources field in the Cluster resource:

apiVersion: k0smotron.io/v1beta1
kind: Cluster
metadata:
  name: k0smotron-test
spec:
  etcd:
    image: quay.io/k0sproject/etcd:v3.5.13
    resources:
      requests:
        cpu: 100m
        memory: 128Mi
      limits:
        cpu: 200m
        memory: 256Mi

This configuration allows you to ensure etcd has appropriate resources allocated, which is especially important for production environments or clusters with high load. The resource requirements follow standard Kubernetes resource specification format.