Fixing Docker Build and Push for Runner Image (#115)

A new image tag for the runner stopped being published on master merges from changes in #86.

This fixes that in the following ways:
- Tests the GH workflow on PRs w/o pushing the images
- Runner and Docker version are moved from Makefile to GH Actions workflow and are passed in as build args
- GHA workflow runs on PRs, and if the workflow file itself is changed (i.e. version bump) or the runner Docker source changes (excluding the Makefile since thats just for local dev)
- Images are pushed on push (i.e. a merge)

.github/workflows/build-runner.yml

@@ -1,14 +1,26 @@
 on:
+  pull_request:
+    branches:
+      - '**'
+    paths:
+      - 'runner/**'
+      - .github/workflows/build-runner.yml
   push:
     branches:
       - master
     paths:
-      - 'runner/**'
-
+      - runner/patched/*
+      - runner/Dockerfile
+      - runner/entrypoint.sh
+      - .github/workflows/build-runner.yml
+name: Runner
 jobs:
   build:
     runs-on: ubuntu-latest
-    name: Build runner
+    name: Build
+    env:
+      RUNNER_VERSION: 2.273.5
+      DOCKER_VERSION: 19.03.12
     steps:
       - name: Checkout
         uses: actions/checkout@v2
@@ -19,16 +31,21 @@ jobs:
         with:
           buildx-version: latest
 
-      - name: Login to GitHub Docker Registry
-        run: echo "${DOCKERHUB_PASSWORD}" | docker login -u "${DOCKERHUB_USERNAME}" --password-stdin
-        env:
-          DOCKERHUB_USERNAME: summerwind
-          DOCKERHUB_PASSWORD: ${{ secrets.DOCKER_ACCESS_TOKEN }}
-
       - name: Build Container Image
         working-directory: runner
         run: |
           docker buildx build \
+            --build-arg RUNNER_VERSION=${RUNNER_VERSION} \
+            --build-arg DOCKER_VERSION=${DOCKER_VERSION} \
             --platform linux/amd64,linux/arm64 \
-            --tag summerwind/actions-runner:latest \
-            -f Dockerfile . --push
+            --tag summerwind/actions-runner:v${RUNNER_VERSION} \
+            -f Dockerfile .
+
+      - name: Push Container Image
+        working-directory: runner
+        run: |
+          docker login -u summerwind --password-stdin <<<${{ secrets.DOCKER_ACCESS_TOKEN }}
+          docker push summerwind/actions-runner:v${RUNNER_VERSION}
+          docker tag summerwind/actions-runner:v${RUNNER_VERSION} summerwind/actions-runner:latest
+          docker push summerwind/actions-runner:latest
+        if: ${{ github.event_name == 'push' }}

.github/workflows/build.yml

@@ -1,45 +0,0 @@
-name: Build
-
-on:
-  push:
-    branches:
-      - master
-    paths-ignore:
-      - 'runner/**'
-      - '.github/**'
-
-jobs:
-  build:
-    runs-on: ubuntu-latest
-    name: Build
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v2
-
-      - name: Install kubebuilder
-        run: |
-          curl -L -O https://github.com/kubernetes-sigs/kubebuilder/releases/download/v2.2.0/kubebuilder_2.2.0_linux_amd64.tar.gz
-          tar zxvf kubebuilder_2.2.0_linux_amd64.tar.gz
-          sudo mv kubebuilder_2.2.0_linux_amd64 /usr/local/kubebuilder
-
-      - name: Run tests
-        run: make test
-
-      - name: Set up Docker Buildx
-        id: buildx
-        uses: crazy-max/ghaction-docker-buildx@v1
-        with:
-          buildx-version: latest
-
-      - name: Login to GitHub Docker Registry
-        run: echo "${DOCKERHUB_PASSWORD}" | docker login -u "${DOCKERHUB_USERNAME}" --password-stdin
-        env:
-          DOCKERHUB_USERNAME: summerwind
-          DOCKERHUB_PASSWORD: ${{ secrets.DOCKER_ACCESS_TOKEN }}
-
-      - name: Build Container Image
-        run: |
-          docker buildx build \
-            --platform linux/amd64,linux/arm64 \
-            --tag summerwind/actions-runner-controller:latest \
-            -f Dockerfile . --push

controllers/autoscaling.go

@@ -4,8 +4,9 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"github.com/summerwind/actions-runner-controller/api/v1alpha1"
 	"strings"
+
+	"github.com/summerwind/actions-runner-controller/api/v1alpha1"
 )
 
 func (r *HorizontalRunnerAutoscalerReconciler) determineDesiredReplicas(rd v1alpha1.RunnerDeployment, hra v1alpha1.HorizontalRunnerAutoscaler) (*int, error) {
@@ -44,6 +45,38 @@ func (r *HorizontalRunnerAutoscalerReconciler) determineDesiredReplicas(rd v1alp
 	}
 
 	var total, inProgress, queued, completed, unknown int
+	type callback func()
+	listWorkflowJobs := func(user string, repoName string, runID int64, fallback_cb callback) {
+		if runID == 0 {
+			fallback_cb()
+			return
+		}
+		jobs, _, err := r.GitHubClient.Actions.ListWorkflowJobs(context.TODO(), user, repoName, runID, nil)
+		if err != nil {
+			r.Log.Error(err, "Error listing workflow jobs")
+			fallback_cb()
+		} else if len(jobs.Jobs) == 0 {
+			fallback_cb()
+		} else {
+			for _, job := range jobs.Jobs {
+				switch job.GetStatus() {
+				case "completed":
+					// We add a case for `completed` so it is not counted in `unknown`.
+					// And we do not increment the counter for completed because
+					// that counter only refers to workflows. The reason for
+					// this is because we do not get a list of jobs for
+					// completed workflows in order to keep the number of API
+					// calls to a minimum.
+				case "in_progress":
+					inProgress++
+				case "queued":
+					queued++
+				default:
+					unknown++
+				}
+			}
+		}
+	}
 
 	for _, repo := range repos {
 		user, repoName := repo[0], repo[1]
@@ -52,20 +85,20 @@ func (r *HorizontalRunnerAutoscalerReconciler) determineDesiredReplicas(rd v1alp
 			return nil, err
 		}
 
-		for _, r := range list.WorkflowRuns {
+		for _, run := range list.WorkflowRuns {
 			total++
 
 			// In May 2020, there are only 3 statuses.
 			// Follow the below links for more details:
 			// - https://developer.github.com/v3/actions/workflow-runs/#list-repository-workflow-runs
 			// - https://developer.github.com/v3/checks/runs/#create-a-check-run
-			switch r.GetStatus() {
+			switch run.GetStatus() {
 			case "completed":
 				completed++
 			case "in_progress":
-				inProgress++
+				listWorkflowJobs(user, repoName, run.GetID(), func() { inProgress++ })
 			case "queued":
-				queued++
+				listWorkflowJobs(user, repoName, run.GetID(), func() { queued++ })
 			default:
 				unknown++
 			}

controllers/autoscaling_test.go

@@ -2,16 +2,17 @@ package controllers
 
 import (
 	"fmt"
+	"net/http/httptest"
+	"net/url"
+	"testing"
+
 	"github.com/summerwind/actions-runner-controller/api/v1alpha1"
 	"github.com/summerwind/actions-runner-controller/github"
 	"github.com/summerwind/actions-runner-controller/github/fake"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
 	clientgoscheme "k8s.io/client-go/kubernetes/scheme"
-	"net/http/httptest"
-	"net/url"
 	"sigs.k8s.io/controller-runtime/pkg/log/zap"
-	"testing"
 )
 
 func newGithubClient(server *httptest.Server) *github.Client {
@@ -44,9 +45,11 @@ func TestDetermineDesiredReplicas_RepositoryRunner(t *testing.T) {
 		sReplicas    *int
 		sTime        *metav1.Time
 		workflowRuns string
+		workflowJobs map[int]string
 		want         int
 		err          string
 	}{
+		// Legacy functionality
 		// 3 demanded, max at 3
 		{
 			repo:         "test/valid",
@@ -122,6 +125,21 @@ func TestDetermineDesiredReplicas_RepositoryRunner(t *testing.T) {
 			workflowRuns: `{"total_count": 4, "workflow_runs":[{"status":"in_progress"}, {"status":"in_progress"}, {"status":"in_progress"}, {"status":"completed"}]}"`,
 			want:         3,
 		},
+
+		// Job-level autoscaling
+		// 5 requested from 3 workflows
+		{
+			repo:         "test/valid",
+			min:          intPtr(2),
+			max:          intPtr(10),
+			workflowRuns: `{"total_count": 4, "workflow_runs":[{"id": 1, "status":"queued"}, {"id": 2, "status":"in_progress"}, {"id": 3, "status":"in_progress"}, {"status":"completed"}]}"`,
+			workflowJobs: map[int]string{
+				1: `{"jobs": [{"status":"queued"}, {"status":"queued"}]}`,
+				2: `{"jobs": [{"status": "in_progress"}, {"status":"completed"}]}`,
+				3: `{"jobs": [{"status": "in_progress"}, {"status":"queued"}]}`,
+			},
+			want: 5,
+		},
 	}
 
 	for i := range testcases {
@@ -136,7 +154,7 @@ func TestDetermineDesiredReplicas_RepositoryRunner(t *testing.T) {
 		_ = v1alpha1.AddToScheme(scheme)
 
 		t.Run(fmt.Sprintf("case %d", i), func(t *testing.T) {
-			server := fake.NewServer(fake.WithListRepositoryWorkflowRunsResponse(200, tc.workflowRuns))
+			server := fake.NewServer(fake.WithListRepositoryWorkflowRunsResponse(200, tc.workflowRuns), fake.WithListWorkflowJobsResponse(200, tc.workflowJobs))
 			defer server.Close()
 			client := newGithubClient(server)
 
@@ -211,6 +229,7 @@ func TestDetermineDesiredReplicas_OrganizationalRunner(t *testing.T) {
 		sReplicas    *int
 		sTime        *metav1.Time
 		workflowRuns string
+		workflowJobs map[int]string
 		want         int
 		err          string
 	}{
@@ -316,6 +335,22 @@ func TestDetermineDesiredReplicas_OrganizationalRunner(t *testing.T) {
 			workflowRuns: `{"total_count": 2, "workflow_runs":[{"status":"in_progress"}, {"status":"completed"}]}"`,
 			err:          "validating autoscaling metrics: spec.autoscaling.metrics[].repositoryNames is required and must have one more more entries for organizational runner deployment",
 		},
+
+		// Job-level autoscaling
+		// 5 requested from 3 workflows
+		{
+			org:          "test",
+			repos:        []string{"valid"},
+			min:          intPtr(2),
+			max:          intPtr(10),
+			workflowRuns: `{"total_count": 4, "workflow_runs":[{"id": 1, "status":"queued"}, {"id": 2, "status":"in_progress"}, {"id": 3, "status":"in_progress"}, {"status":"completed"}]}"`,
+			workflowJobs: map[int]string{
+				1: `{"jobs": [{"status":"queued"}, {"status":"queued"}]}`,
+				2: `{"jobs": [{"status": "in_progress"}, {"status":"completed"}]}`,
+				3: `{"jobs": [{"status": "in_progress"}, {"status":"queued"}]}`,
+			},
+			want: 5,
+		},
 	}
 
 	for i := range testcases {
@@ -330,7 +365,7 @@ func TestDetermineDesiredReplicas_OrganizationalRunner(t *testing.T) {
 		_ = v1alpha1.AddToScheme(scheme)
 
 		t.Run(fmt.Sprintf("case %d", i), func(t *testing.T) {
-			server := fake.NewServer(fake.WithListRepositoryWorkflowRunsResponse(200, tc.workflowRuns))
+			server := fake.NewServer(fake.WithListRepositoryWorkflowRunsResponse(200, tc.workflowRuns), fake.WithListWorkflowJobsResponse(200, tc.workflowJobs))
 			defer server.Close()
 			client := newGithubClient(server)
 

github/fake/fake.go

@@ -4,7 +4,10 @@ import (
 	"fmt"
 	"net/http"
 	"net/http/httptest"
+	"strconv"
+	"strings"
 	"time"
+	"unicode"
 )
 
 const (
@@ -31,6 +34,24 @@ func (h *Handler) ServeHTTP(w http.ResponseWriter, req *http.Request) {
 	fmt.Fprintf(w, h.Body)
 }
 
+type MapHandler struct {
+	Status int
+	Bodies map[int]string
+}
+
+func (h *MapHandler) ServeHTTP(w http.ResponseWriter, req *http.Request) {
+	// Parse out int key from URL path
+	key, err := strconv.Atoi(strings.TrimFunc(req.URL.Path, func(r rune) bool { return !unicode.IsNumber(r) }))
+	if err != nil {
+		w.WriteHeader(400)
+	} else if body := h.Bodies[key]; len(body) == 0 {
+		w.WriteHeader(404)
+	} else {
+		w.WriteHeader(h.Status)
+		fmt.Fprintf(w, body)
+	}
+}
+
 type ServerConfig struct {
 	*FixedResponses
 }
@@ -45,7 +66,7 @@ func NewServer(opts ...Option) *httptest.Server {
 		o(&config)
 	}
 
-	routes := map[string]*Handler{
+	routes := map[string]http.Handler{
 		// For CreateRegistrationToken
 		"/repos/test/valid/actions/runners/registration-token": &Handler{
 			Status: http.StatusCreated,
@@ -126,6 +147,9 @@ func NewServer(opts ...Option) *httptest.Server {
 
 		// For auto-scaling based on the number of queued(pending) workflow runs
 		"/repos/test/valid/actions/runs": config.FixedResponses.ListRepositoryWorkflowRuns,
+
+		// For auto-scaling based on the number of queued(pending) workflow jobs
+		"/repos/test/valid/actions/runs/": config.FixedResponses.ListWorkflowJobs,
 	}
 
 	mux := http.NewServeMux()

github/fake/options.go

@@ -2,6 +2,7 @@ package fake
 
 type FixedResponses struct {
 	ListRepositoryWorkflowRuns *Handler
+	ListWorkflowJobs           *MapHandler
 }
 
 type Option func(*ServerConfig)
@@ -15,6 +16,15 @@ func WithListRepositoryWorkflowRunsResponse(status int, body string) Option {
 	}
 }
 
+func WithListWorkflowJobsResponse(status int, bodies map[int]string) Option {
+	return func(c *ServerConfig) {
+		c.FixedResponses.ListWorkflowJobs = &MapHandler{
+			Status: status,
+			Bodies: bodies,
+		}
+	}
+}
+
 func WithFixedResponses(responses *FixedResponses) Option {
 	return func(c *ServerConfig) {
 		c.FixedResponses = responses

runner/Makefile

@@ -1,7 +1,7 @@
 NAME ?= summerwind/actions-runner
 TAG ?= latest
 
-RUNNER_VERSION ?= 2.273.4
+RUNNER_VERSION ?= 2.273.5
 DOCKER_VERSION ?= 19.03.12
 
 # default list of platforms for which multiarch image is built