The configuration for proxy-injection and the linkerd-config configmap

[krishgu]

Team,

• when linkerd injects the proxy into a pod, where does it get the proxy's configuration from? for example

  • cpu/memory resource requests/limits
  • environment variables such as LINKERD2_PROXY_LOG
  • imageregistry?

• I always thought that this was from the configmap linkerd-config, if that's not the case - is the configmap only used for addons? what is the minimal configmap to make a working version of linkerd?

  • does the configmap need to be kustomized for each environment?

• Looked at discussion 4966 and the PR 4915 Remove dependency of linkerd-config for control plane components #4915 - but it is not clear where these configs are passed as command-line arguments in the proxy-injector.

Thanks!

[krishgu]

Using stable-2.14.10, I tested this by modifying linkerd-config configmap, and restarting the proxy-injector deployment. The sections here from the proxy section seem to take effect.

We can now use kustomize to patch the configmap, to kustomize registries/resources for each cluster. (replacing the full values)

    proxy:
     ...
      enableExternalProfiles: false
      image:
        name: myregistry.etc/linkerd/proxy
        pullPolicy: ""
        version: ""
     ...
      opaquePorts: 25,587,3306,4444,5432,6379,9300,11211
    ...
      resources:
        cpu:
          limit: ""
          request: 27m
        memory:
          limit: 199Mi
          request: 23Mi