How limit worker resource ?

[0Delta]

I want limit worker resource for simulate weak node. e.g. 1vCPU/2GB

• kind version 0.8.1
• Windows 10 2004 build:19041.208
• WSL2 with Ubuntu 18.04
  • Linux MyPC 4.19.84-microsoft-standard Add all the things: port history from test-infra #1 SMP Wed Nov 13 11:44:37 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
• Docker Desktop 2.2.0.4 (43472)
  • Engine 19.03.8
  • Kubernetes v1.15.5
  • Compose 1.25.4
  • Notary 0.6.1
  • Credential Helper 0.6.3

I tryed ...

• using constraints parameter, such as Configure capacity of the worker nodes #877
  but, i got syntax error.

Creating cluster "kind" ...
ERROR: failed to create cluster: unable to decode config: yaml: unmarshal errors:
  line 7: field constraints not found in type v1alpha3.Node

shell returned 1

• using kubeadmConfigPatches parameter, such as How to set Node Allocatable memory in kind? #1524
  but it's look no effects.

$ kubectl describe nodes
 : 
Addresses:
  InternalIP:  172.20.0.2
  Hostname:    kind-control-plane
Capacity:
 cpu:                12                              # I want restrict this parameter
 ephemeral-storage:  263174212Ki
 hugepages-2Mi:      0
 memory:             25912152Ki           # and this one as well.
 pods:               110
Allocatable:
 cpu:                12
 ephemeral-storage:  263174212Ki
 hugepages-2Mi:      0
 memory:             21717848Ki
 pods:               110
 : 

any ideas ?

[BenTheElder]

#1524
#877

If you use an older kubernetes version you'll have to use older kubeadm config types.

[teachermanalex]

@0Delta, @BenTheElder :
I found that I can use kubeadmConfigPatches and it will affect the control-plane node but not the worker node....

[teachermanalex]

Ok, figured it out. For worker nodes, you have to use JoinConfiguration:

- role: worker
  kubeadmConfigPatches:
  - |
    kind: JoinConfiguration
    nodeRegistration:
      kubeletExtraArgs:
        system-reserved: memory=8Gi

[aojea]

I didn't test it but you should be able to patch the kubelet configuration directly

---
apiVersion: kubelet.config.k8s.io/v1beta1
kind: KubeletConfiguration
metadata:
  name: config
systemReserved:
  cpu: 2
  memory: "8Gi"

[BenTheElder]

kubelet configuration object is not respected per node in kubeadm currently, only from init :(

this is what you want #1578 (comment)

[0Delta]

Sorry for late response.

finally, I success controle resource by means of @teachermanalex .

- role: worker
  kubeadmConfigPatches:
  - |
    kind: JoinConfiguration
    nodeRegistration:
      kubeletExtraArgs:
        system-reserved: memory=8Gi,cpu=8

And i found this issue. #896
that mean "resource controle function is now developing. but it not matured, so it too early to merge."
I will use kubeadmConfigPatches carefully. becase it is compromised method.

However, the control node resource is successful. my question is solved.
thanks for all.

[Al4DIN]

Hello, can we specify the ram utilisation with an argument instead of passing it in our yaml file.
For example kubectl cluster-info --context kind-kind -arg

[BenTheElder]

Hello, can we specify the ram utilisation with an argument instead of passing it in our yaml file.

Most complex options are only in config, not a flag, and will remain that way.
This allows us to version the types, behavior, etc.
Kubernetes core components also work this way, it is referred to as "component config" there.