[installer] Allow configuration of resource requests and limits

[andrew-farries]

Description

One of the Webapp team's epics for Q2 is to use the Gitpod installer to deploy to Gitpod SaaS. In order to do that we will need to add additional configuration to the installer to make the output suitable for a SaaS deployment as opposed to a self-hosted deployment.

This PR adds the ability to configure the resource requests and limits for all components via the experimental section in the installer config file.

Related Issue(s)

Part of #9097

How to test

Create an installer config file containing this experimental section:

experimental:
  common:
    podConfig:
      server:
        resources:
          server:
            requests:
              cpu: 50m
              memory: 100Mi
            limits:
              cpu: 500m
              memory: 800Mi
      dashboard:
        resources:
          dashboard:
            requests:
              cpu: 60m
              memory: 100Mi
            limits:
              cpu: 100m
              memory: 500Mi

Get a versions.yaml for use with the installer:

docker run -it --rm "eu.gcr.io/gitpod-core-dev/build/versions:${version}" cat versions.yaml > versions.yaml

Then invoke the installer as:

go run . render --debug-version-file versions.yaml --config /path/to/config --use-experimental-config

The rendered output will have the values from the config for the server and dashboard component resource requests and limits. All other components will use their default hard-coded resources as before.

Release Notes

Allow resource requests and limits for each component to be configurable through the installer

Documentation

None.

[andrew-farries]

I'm not sure whether it's worth making the kube-rbac-proxy component that's used in multiple pods also configurable:

gitpod/install/installer/pkg/common/common.go

Lines 305 to 308 in 3998444

	Resources: corev1.ResourceRequirements{Requests: corev1.ResourceList{
	corev1.ResourceCPU: resource.MustParse("1m"),
	corev1.ResourceMemory: resource.MustParse("30Mi"),
	}},

[geropl]

I'm not sure whether it's worth making the kube-rbac-proxy component that's used in multiple pods also configurable:

@andrew-farries ah, good find! No, let's keep that as is for now. 👍

[geropl]

❤️

[geropl]

LGTM, and tests run! 👍

[andrew-farries]

/werft run

👍 started the job as gitpod-build-af-installer-specify-resourcerequirements.4
(with .werft/ from main)

[mustard-mh]

incorrectServerResources := corev1.ResourceRequirements{
	Requests: corev1.ResourceList{
		"cpu":    resource.MustParse("0"),
		"memory": resource.MustParse("0"),
	},
	Limits: corev1.ResourceList{
		"cpu":    resource.MustParse("-500m"),
		"memory": resource.MustParse("800Mi"),
	},
}

after common.ResourceRequirements func, output goes

{
        "limits": {
                "cpu": "-500m",
                "memory": "800Mi"
        },
        "requests": {
                "cpu": "0",
                "memory": "0"
        }
}

Should we validate it and panic first? or let it failed after apply to k8s

[geropl]

Should we validate it and panic first? or let it failed after apply to k8s

@mustard-mh IMO that's a kubernetes thing, as we're re-using their data structures. 👍