Default Resource Requests and Limits overwrites empty/null value

[felfa01]

Describe the problem/challenge you have
You are not able to configure null values for restic daemonset or velero deployment containers. If left empty or with null/~ it will always use the default values. For some use cases you do not want to have e.g. CPU Limits.

Describe the solution you'd like
Allow empty values for CPU/Memory Requests/Limits.

Environment:

• helm version (use helm version): version.BuildInfo{Version:"v3.6.2", GitCommit:"ee407bdf364942bcb8e8c665f82e15aa28009b71", GitTreeState:"clean", GoVersion:"go1.16.5"}
• helm chart version and app version (use helm list -n <YOUR NAMESPACE>): 2.23.10
• Kubernetes version (use kubectl version):v1.21.9
• Kubernetes installer & version:
• Cloud provider or hardware configuration: azure
• OS (e.g. from /etc/os-release):

[jenting]

@felfa01 I did a test with chart 2.28.0 by changing the resources from

resources:
  requests:
    cpu: 500m
    memory: 512Mi
  limits:
    cpu: 1000m
    memory: 1024Mi

to

resources: {}

The rendered manifest without the cpu/memory requests/limits.

Also, I configured resources.requests only (without limits), the rendered manifest is correct.

resources:
  requests:
    cpu: 500m
    memory: 512Mi

could you please check it?

[felfa01]

@jenting Yes you are correct that what you do works. However, if you are passing a value files with -f flag into e.g. helm upgrade that looks like this:

resources:
  requests:
    memory: 512Mi
  limits:
    cpu: 1000m
    memory: 1024Mi

or

resources:
  requests:
    cpu: 500m
    memory: 512Mi
  limits:
    cpu: ~
    memory: null

all empty or null/~ will be overwritten by the default value file from charts/velero/values.yaml.

[jenting]

@felfa01 Please try this

resources:
  requests:
    cpu: null
    memory: 512Mi
  limits:
    cpu: 1000m
    memory: 1024Mi

or

resources:
  requests:
    cpu: 500m
    memory: 512Mi
  limits: null

I use the helm v3.7.1 for testing.

[felfa01]

@jenting
Sorry for not being more clear. What I have setup is a dependency to this repo:

dependencies:
  - name: velero
    version: 2.23.10
    repository: https://vmware-tanzu.github.io/helm-charts

If I in my sub-chart have a values.yaml and configure the following:

#values.yaml
  resources:
    requests:
      cpu: null
      memory: 1Gi
    limits:
      cpu: 1
      memory: 1Gi

and run helm template I get the following output:

  resources:
    requests:
      cpu: 500m
      memory: 1Gi
    limits:
      cpu: 1
      memory: 1Gi

However, if I add an additional test-values.yaml as input looking like this:

#test-values.yaml
  resources: 
    requests:
      cpu: null
      memory: null

and run helm template charts/ --values test-values.yaml I get this:

  resources:
    requests:
      memory: 1Gi
    limits:
      cpu: 1
      memory: 1Gi

I get the right value for requests.cpu but not for requests.memory.
My troubleshooting seems to point to some issue with chart dependencies. Something similar can be mentioned here: helm/helm#5184 (comment)

[jenting]

I see.

It's the helm issue with sub-chart not Velero helm chart issue.
Therefore, your currently PR is a workaround way.

[jenting]

I'll close this issue because it's not Velero helm chart issue, but a helm issue.
Feel free to reopen it if you still have concerns.