Increase resource allocation to runner pod

Grafana k6 k6-operator
1

Hi, I am trying to increase the resource allocation to the runner pod as the test cases that are executed are pretty big and the runner pods keep getting OOMKilled.

image
image1347×97 22.4 KB

When I described the runner pod, I see the pod had been using the default settings even though I increased the resource allocation.

apiVersion: k6.io/v1alpha1
kind: TestRun
metadata:
  name: k6
spec:
  parallelism: 1
  cleanup: post
  arguments: -o experimental-prometheus-rw --tag testid=call-report_2
  script:
    configMap:
      name: demo
      file: archive.tar
  runner:
    env:
      - name: K6_PROMETHEUS_RW_SERVER_URL
        value: "http://prometheus-kube-prometheus-prometheus.monitor.svc.cluster.local:9090/api/v1/write"
      - name: K6_PROMETHEUS_RW_TREND_AS_NATIVE_HISTOGRAM
        value: "true"
    resources:
      limits:
        cpu: "3000m"
        memory: "25Gi"
      requests:
        cpu: "2000m"
        memory: "25Gi"

Below is the resources that were actually allocated to the runner pod. Why wasn’t the increase of resources applied?
image

2

Hi @zzhao2022

It is wierd, because it should work like that.
For me this works:

apiVersion: k6.io/v1alpha1
kind: TestRun
metadata:
  name: perf-test
spec:
  parallelism: 1
  script:
    configMap:
      name: perf-test
      file: perf-test.js
  runner:
    resources:
      limits:
        cpu: "300m"
        memory: "250M"
      requests:
        cpu: "300m"
        memory: "250M"

In the descriptor of the created runner pod I can see:

...
    resources:
      limits:
        cpu: 300m
        memory: 250M
      requests:
        cpu: 300m
        memory: 250M
...

Did you delete the TestRun resource before updating it?

4

Hi @bandorko,
I am not sure what was the fix. I re-deployed k6-operator from the beginning. And it works now. I see the custom resource allocation settings are applied to the runner pod correctly now.

I have a follow-up question. The test case I plan to run via k6-operator is quite large. It could consume more than 25 GB mem when I was testing on my local because the test case contains multiple scenarios. So in this case, besides increasing resources for the runner pod, do I also need to increase the resources for the manager pod, starter pod and initiate pod? And is there any resource limit? For example, is there any issue if I request like 25 GiB of memory for the runner pod?
Is there any recommendations for running big test case? I don’t see any in the documentation.