next

Edit

Pod Memory Hog Details

Experiment Metadata

Type Description Tested K8s Platform
Generic Consume memory resources on the application container GKE, Packet(Kubeadm), Minikube, EKS, AKS

Prerequisites

  • Ensure that Kubernetes Version > 1.16
  • Ensure that the Litmus Chaos Operator is running by executing kubectl get pods in operator namespace (typically, litmus). If not, install from here
  • Ensure that the pod-memory-hog experiment resource is available in the cluster by executing kubectl get chaosexperiments in the desired namespace. If not, install from here
  • Cluster must run docker container runtime

Entry Criteria

  • Application pods are healthy on the respective nodes before chaos injection

Exit Criteria

  • Application pods are healthy on the respective nodes post chaos injection

Details

  • This experiment consumes the Memory resources on the application container on specified memory in megabytes.
  • It simulates conditions where app pods experience Memory spikes either due to expected/undesired processes thereby testing how the overall application stack behaves when this occurs.

Integrations

  • Pod Memory can be effected using the chaos library: litmus

Steps to Execute the Chaos Experiment

  • This Chaos Experiment can be triggered by creating a ChaosEngine resource on the cluster. To understand the values to provide in a ChaosEngine specification, refer Getting Started

  • Follow the steps in the sections below to create the chaosServiceAccount, prepare the ChaosEngine & execute the experiment.

Prepare chaosServiceAccount

Use this sample RBAC manifest to create a chaosServiceAccount in the desired (app) namespace. This example consists of the minimum necessary role permissions to execute the experiment.

Sample Rbac Manifest

---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: pod-memory-hog-sa
  namespace: default
  labels:
    name: pod-memory-hog-sa
    app.kubernetes.io/part-of: litmus
---
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: pod-memory-hog-sa
  namespace: default
  labels:
    name: pod-memory-hog-sa
    app.kubernetes.io/part-of: litmus
rules:
- apiGroups: [""]
  resources: ["pods","events"]
  verbs: ["create","list","get","patch","update","delete","deletecollection"]
- apiGroups: [""]
  resources: ["pods/exec","pods/log","replicationcontrollers"]
  verbs: ["create","list","get"]
- apiGroups: ["batch"]
  resources: ["jobs"]
  verbs: ["create","list","get","delete","deletecollection"]
- apiGroups: ["apps"]
  resources: ["deployments","statefulsets","daemonsets","replicasets"]
  verbs: ["list","get"]
- apiGroups: ["apps.openshift.io"]
  resources: ["deploymentconfigs"]
  verbs: ["list","get"]
- apiGroups: ["argoproj.io"]
  resources: ["rollouts"]
  verbs: ["list","get"]
- apiGroups: ["litmuschaos.io"]
  resources: ["chaosengines","chaosexperiments","chaosresults"]
  verbs: ["create","list","get","patch","update"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: pod-memory-hog-sa
  namespace: default
  labels:
    name: pod-memory-hog-sa
    app.kubernetes.io/part-of: litmus
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: pod-memory-hog-sa
subjects:
- kind: ServiceAccount
  name: pod-memory-hog-sa
  namespace: default

Note: In case of restricted systems/setup, create a PodSecurityPolicy(psp) with the required permissions. The chaosServiceAccount can subscribe to work around the respective limitations. An example of a standard psp that can be used for litmus chaos experiments can be found here.

Prepare ChaosEngine

  • Provide the application info in spec.appinfo
  • Override the experiment tunables if desired in experiments.spec.components.env
  • To understand the values to provided in a ChaosEngine specification, refer ChaosEngine Concepts

Supported Experiment Tunables

Variables Description Type Notes
MEMORY_CONSUMPTION The amount of memory used of hogging a Kubernetes pod (megabytes) Optional Defaults to 500MB (Up to 2000MB)
NUMBER_OF_WORKERS The number of workers used to run the stress process Optional Defaults to 1
TOTAL_CHAOS_DURATION The time duration for chaos insertion (seconds) Optional Defaults to 60s
LIB The chaos lib used to inject the chaos. Available libs are litmus and pumba Optional Defaults to litmus
LIB_IMAGE Image used to run the helper pod. Optional Defaults to litmuschaos/go-runner:1.13.8
STRESS_IMAGE Container run on the node at runtime by the pumba lib to inject stressors. Only used in LIB pumba Optional Default to alexeiled/stress-ng:latest-ubuntu
TARGET_PODS Comma separated list of application pod name subjected to pod memory hog chaos Optional If not provided, it will select target pods randomly based on provided appLabels
TARGET_CONTAINER Name of the target container under chaos. Optional If not provided, it will select the first container of the target pod
CONTAINER_RUNTIME container runtime interface for the cluster Optional Defaults to docker, supported values: docker, containerd and crio for litmus and only docker for pumba LIB
SOCKET_PATH Path of the containerd/crio/docker socket file Optional Defaults to `/var/run/docker.sock`
PODS_AFFECTED_PERC The Percentage of total pods to target Optional Defaults to 0 (corresponds to 1 replica), provide numeric value only
RAMP_TIME Period to wait before and after injection of chaos in sec Optional
SEQUENCE It defines sequence of chaos execution for multiple target pods Optional Default value: parallel. Supported: serial, parallel
INSTANCE_ID A user-defined string that holds metadata/info about current run/instance of chaos. Ex: 04-05-2020-9-00. This string is appended as suffix in the chaosresult CR name. Optional Ensure that the overall length of the chaosresult CR is still < 64 characters

Sample ChaosEngine Manifest

apiVersion: litmuschaos.io/v1alpha1
kind: ChaosEngine
metadata:
  name: nginx-chaos
  namespace: default
spec:
  # It can be active/stop
  engineState: 'active'
  appinfo:
    appns: 'default'
    applabel: 'app=nginx'
    appkind: 'deployment'
  chaosServiceAccount: pod-memory-hog-sa
  experiments:
    - name: pod-memory-hog
      spec:
        components:
          env:
            - name: TOTAL_CHAOS_DURATION
              value: '60' # in seconds

            # Enter the amount of memory in megabytes to be consumed by the application pod
            - name: MEMORY_CONSUMPTION
              value: '500'

             ## percentage of total pods to target
            - name: PODS_AFFECTED_PERC
              value: ''   

            ## provide the cluster runtime
            - name: CONTAINER_RUNTIME
              value: 'docker'   

            # provide the socket file path
            - name: SOCKET_PATH
              value: '/var/run/docker.sock'

Create the ChaosEngine Resource

  • Create the ChaosEngine manifest prepared in the previous step to trigger the Chaos.

    kubectl apply -f chaosengine.yml

  • If the chaos experiment is not executed, refer to the troubleshooting section to identify the root cause and fix the issues.

Watch Chaos progress

  • Set up a watch on the applications interacting/dependent on the affected pods and verify whether they are running

    watch kubectl get pods -n <application-namespace>

Abort/Restart the Chaos Experiment

  • To stop the pod-memory-hog experiment immediately, either delete the ChaosEngine resource or execute the following command:

    kubectl patch chaosengine <chaosengine-name> -n <namespace> --type merge --patch '{"spec":{"engineState":"stop"}}'

  • To restart the experiment, either re-apply the ChaosEngine YAML or execute the following command:

    kubectl patch chaosengine <chaosengine-name> -n <namespace> --type merge --patch '{"spec":{"engineState":"active"}}'

Check Chaos Experiment Result

  • Check whether the application stack is resilient to Memory spikes on the app replica, once the experiment (job) is completed. The ChaosResult resource name is derived like this: <ChaosEngine-Name>-<ChaosExperiment-Name>.

    kubectl describe chaosresult nginx-chaos-pod-memory-hog -n <application-namespace>

Pod Memory Hog Experiment Demo

  • A sample recording of this experiment execution is provided here
← Pod CPU HogPod DNS Error →