Skip to content

Latest commit

 

History

History
66 lines (52 loc) · 3.24 KB

README.md

File metadata and controls

66 lines (52 loc) · 3.24 KB
Raw

Resource Quotas (CPU & Memory)

This check tests if namespace resource quotas CPU and memory are under a specified threshold or percentage. Namespaces that utilize a lot of CPU or memory resources can sometimes run into an issue where controllers (i.e. deployment or replica controllers) are unable to schedule pods due to insufficient CPU or memory.

This check lists all namespaces in the cluster and checks if each resource (CPU and memory) are at an ok percentage.

This check can be configured to use either a blacklist or a whitelist of namespaces, allowing you to explicitly target or ignore specific namespaces. If any namespaces for the check need to be on the blacklist or whitelist they can be specified with the environment variables BLACKLIST and WHITELIST which expect a comma-separated list of namespaces ("default,kube-system,istio-system") and can help you configure which namespaces to check when used in combination, with the BLACKLIST and WHITELIST environment variables.

Additionally, a threshold or percentage can be set that will determine when the check will configure and create alert messages. You can configure this value with the environment variable THRESHOLD, which expects a float value between 0.0 and 1.00 (not inclusive). By default, the threshold is set to 0.90 or 90%

Check Steps

This check follows the list of actions in order during the run of the check:

  1. Lists all namespaces in the cluster.
  2. Sends a go routine for each namespace.
  3. Each go routine checks if used CPU and memory have reached the threshold.
  4. Each go routine creates errors for each violating namespace. (Up to two errors -- one for CPU and one for memory)

Check Details

  • Namespace: kuberhealthy
  • Check name: resource-quota
  • Configurable check environment variables:
    • BLACKLIST: Blacklist of namespaces to look at (default for BLACKLIST=default)
    • WHITELIST: Whitelist of namespaces to look at. (default for whitelist=kube-system,kuberhealthy)
    • THRESHOLD: Percentage or threshold for usage that should determine whether or not an error should be created. Expects a float value. (default=0.9)
    • DEBUG: Turns on debug logging. (default=false)

Example KuberhealthyCheck Spec

---
apiVersion: comcast.github.io/v1
kind: KuberhealthyCheck
metadata:
  name: resource-quota
  namespace: kuberhealthy
spec:
  runInterval: 1h
  timeout: 2m
  podSpec:
    containers:
    - name: resource-quota
      image: kuberhealthy/resource-quota-check:v1.3.0
      imagePullPolicy: IfNotPresent
      env:
        - name: BLACKLIST
          value: "default"
        - name: WHITELIST
          value: "kube-system,kuberhealthy"
      resources:
        requests:
          cpu: 15m
          memory: 15Mi
        limits:
          cpu: 30m
    restartPolicy: Never
    terminationGracePeriodSeconds: 30

Install

To use the Resource Quota Check with Kuberhealthy, apply the configuration file resource-quota.yaml to your Kubernetes Cluster. The following command will also apply the configuration file to your current context:

kubectl apply -f https://raw.githubusercontent.com/kuberhealthy/kuberhealthy/cmd/resource-quota-check/resource-quota-check.yaml