Downward API implementation for resources limits and requests

[aveshagarwal]

This is an implementation of Downward API for resources limits and requests, and it works with environment variables and volume plugin.

This is based on proposal #24051. This implementation follows API with magic keys approach as discussed in the proposal.

@kubernetes/rh-cluster-infra

---

This change is

[aveshagarwal]

Updated PR. Will add more tests soon.
@bgrant0607 @derekwaynecarr @pmorie PTAL.

[aveshagarwal]

Added some validation tests and fixed some issues.

[aveshagarwal]

I have tested it with following in case any one is interested:

Env variables:

apiVersion: v1
kind: Pod
metadata:
  name: dapi-test-pod
spec:
  containers:
    - name: test-container
      image: gcr.io/google_containers/busybox
      command: [ "/bin/sh", "-c", "env" ]
      resources:
        requests:
          memory: "64Mi"
          cpu: "250m"
        limits:
          memory: "128Mi"
          cpu: "1500m"
      env:
        - name: CPU_LIMIT
          valueFrom:
            resourceFieldRef:
              resource: limits.cpu
    - name: test-container2
      image: gcr.io/google_containers/busybox
      command: [ "/bin/sh", "-c", "env" ]
      resources:
        requests:
          memory: "32Mi"
          cpu: "125m"
        limits:
          memory: "64Mi"
          cpu: "250m"
      env:
        - name: MEMORY_LIMIT
          valueFrom:
            resourceFieldRef:
              resource: limits.memory
        - name: CPU_LIMIT
          valueFrom:
            resourceFieldRef:
              resource: requests.cpu
  restartPolicy: Never

Volume plugin:

apiVersion: v1
kind: Pod
metadata:
  name: kubernetes-downwardapi-volume-example
  labels:
    zone: us-est-coast
    cluster: test-cluster1
    rack: rack-22
  annotations:
    build: two
    builder: john-doe
spec:
  containers:
    - name: client-container
      image: gcr.io/google_containers/busybox
      command: ["sh", "-c", "while true; do if [[ -e /etc/labels ]]; then cat /etc/labels; fi; if [[ -e /etc/annotations ]]; then cat /etc/annotations; fi; sleep 5; done"]
      resources:
        requests:
          memory: "32Mi"
          cpu: "125m"
        limits:
          memory: "64Mi"
          cpu: "250m"
      volumeMounts:
        - name: podinfo
          mountPath: /etc
          readOnly: false
  volumes:
    - name: podinfo
      downwardAPI:
        items:
          - path: "labels"
            fieldRef:
              fieldPath: metadata.labels
          - path: "annotations"
            fieldRef:
              fieldPath: metadata.annotations
          - path: "client-container/pod_name"
            fieldRef:
              fieldPath: metadata.name
          - path: "cpu_limit"
            resourceFieldRef:
              containerName: client-container
              resource: limits.cpu
              divisor: "1"
          - path: "cpu_request"
            resourceFieldRef:
              containerName: client-container
              resource: requests.cpu
              divisor: "1m"
          - path: "memory_limit"
            resourceFieldRef:
              containerName: client-container
              resource: limits.memory
              divisor: "1M"
          - path: "memory_request"
            resourceFieldRef:
              containerName: client-container
              resource: requests.memory

[aveshagarwal]

@bgrant0607 @derekwaynecarr @pmorie I have tested it with the above configs and also some other variations including error conditions.

[pmorie]

Add info about the default here.

[pmorie]

@aveshagarwal this is moving in the right direction. Please add e2es for envs and volumes.

[aveshagarwal]

@bgrant0607 @derekwaynecarr @pmorie I have updated this PR.

1. Added e2e tests for env vars, and tested locally and they work.
2. Improved validation for memory divisor.
3. Addressed pmorie's comments.

Still working on e2e for volume.

[aveshagarwal]

Added e2e tests for volumes too now and tested locally to make sure they work.

@bgrant0607 @pmorie @derekwaynecarr PTAL.

[pmorie]

s/.Child("path")//

@aveshagarwal

[pmorie]

@aveshagarwal I'll take the nits in a follow-up. LGTM. Thanks a lot.

[aveshagarwal]

I see a failure in unit tests: FAIL k8s.io/kubernetes/pkg/api , it seems to be failing in pkg/api/serialization_test.go.

[aveshagarwal]

I regenerated auto-generated files and when I test ./hack/test-go.sh pkg/api now it fails in pkg/api/copy_test.go. Not sure what is happening.

[aveshagarwal]

I have found the solution and will just run the tests locally and update it and hopefully it should be good to go.

[pmorie]

I will re-tag once green

[aveshagarwal]

This e2e failure does not seem related to this PR:

[Fail] [k8s.io] Pods [It] should invoke init containers on a RestartNever pod 
/go/src/k8s.io/kubernetes/_output/dockerized/go/src/k8s.io/kubernetes/test/e2e/pods.go:710

[aveshagarwal]

Seems similar to this: #26131

[pmorie]

@k8s-bot test this issue #26131

[aveshagarwal]

@pmorie all green, LGTM again?

[aveshagarwal]

@bgrant0607 @derekwaynecarr can you tag it LGTM again?

[k8s-github-robot]

@k8s-bot test this [submit-queue is verifying that this PR is safe to merge]

[k8s-bot]

GCE e2e build/test passed for commit 1931931.

• Build Log
• Test Artifacts
• Internal Jenkins Results

[k8s-github-robot]

Automatic merge from submit-queue