pod-memory-stress

Allocates memory inside the pod's cgroup, simulating memory pressure and testing OOM behavior.

Target kind: Pod
Implementation: Daemon (ExecStressChaos with stressorType: memory)
Rollback: Yes — sends CancelChaos to terminate stress-ng

Parameters

Name	Type	Required	Default	Description
workers	string	No	"1"	Number of memory stress worker threads
size	string	No	"256m"	Memory to allocate per worker (e.g. 256m, 1g)
duration	string	No	experiment duration	How long to run stress

Example

apiVersion: chaos.chaosplane.io/v1alpha1
kind: ChaosExperiment
metadata:
  name: pod-memory-stress-example
  namespace: default
spec:
  target:
    kind: Pod
    namespace: default
    labelSelector:
      matchLabels:
        app: worker
  action:
    type: pod-memory-stress
    parameters:
      workers: "1"
      size: "512m"
      duration: "60s"
  duration: 60s
  rollback:
    enabled: true

Rollback behavior

Sends CancelChaos RPC to the daemon, which terminates the stress-ng process and releases the allocated memory.

Implementation notes

The daemon runs stress-ng --vm N --vm-bytes SIZE inside the pod's cgroup. Memory allocation is subject to the pod's memory limits — if the pod has a resources.limits.memory set, the stress may trigger an OOM kill. This is intentional for testing OOM recovery.

Use abort conditions with a k8s probe to detect if the pod gets OOM-killed and abort the experiment.