runner-controller: do not delete runner if it is busy (#103)

Currently, after refreshing the token, the controller re-creates the runner with the new token. This results in jobs being interrupted. This PR makes sure the pod is not restarted if it is busy.

Closes #74
This commit is contained in:
Helder Moreira
2020-10-05 01:06:37 +01:00
committed by GitHub
parent 7b5e62e266
commit 7a2fa7fbce
16 changed files with 539 additions and 102 deletions

View File

@@ -2,9 +2,10 @@ package controllers
import (
"context"
"github.com/summerwind/actions-runner-controller/github/fake"
"time"
"github.com/summerwind/actions-runner-controller/github/fake"
corev1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/types"
"k8s.io/client-go/kubernetes/scheme"
@@ -44,7 +45,7 @@ func SetupIntegrationTest(ctx context.Context) *testEnvironment {
Status: 200,
Body: workflowRunsFor3Replicas,
}
server := fake.NewServer(fake.WithFixedResponses(responses))
fakeGithubServer := fake.NewServer(fake.WithFixedResponses(responses))
BeforeEach(func() {
stopCh = make(chan struct{})
@@ -58,11 +59,16 @@ func SetupIntegrationTest(ctx context.Context) *testEnvironment {
mgr, err := ctrl.NewManager(cfg, ctrl.Options{})
Expect(err).NotTo(HaveOccurred(), "failed to create manager")
runnersList = fake.NewRunnersList()
server = runnersList.GetServer()
ghClient := newGithubClient(server)
replicasetController := &RunnerReplicaSetReconciler{
Client: mgr.GetClient(),
Scheme: scheme.Scheme,
Log: logf.Log,
Recorder: mgr.GetEventRecorderFor("runnerreplicaset-controller"),
Client: mgr.GetClient(),
Scheme: scheme.Scheme,
Log: logf.Log,
Recorder: mgr.GetEventRecorderFor("runnerreplicaset-controller"),
GitHubClient: ghClient,
}
err = replicasetController.SetupWithManager(mgr)
Expect(err).NotTo(HaveOccurred(), "failed to setup controller")
@@ -76,7 +82,7 @@ func SetupIntegrationTest(ctx context.Context) *testEnvironment {
err = deploymentsController.SetupWithManager(mgr)
Expect(err).NotTo(HaveOccurred(), "failed to setup controller")
client := newGithubClient(server)
client := newGithubClient(fakeGithubServer)
autoscalerController := &HorizontalRunnerAutoscalerReconciler{
Client: mgr.GetClient(),
@@ -99,7 +105,7 @@ func SetupIntegrationTest(ctx context.Context) *testEnvironment {
AfterEach(func() {
close(stopCh)
server.Close()
fakeGithubServer.Close()
err := k8sClient.Delete(ctx, ns)
Expect(err).NotTo(HaveOccurred(), "failed to delete test namespace")

View File

@@ -203,12 +203,16 @@ func (r *RunnerReconciler) Reconcile(req ctrl.Request) (ctrl.Result, error) {
return ctrl.Result{}, err
}
if pod.Spec.Containers[0].Image != newPod.Spec.Containers[0].Image {
restart = true
}
if !reflect.DeepEqual(pod.Spec.Containers[0].Env, newPod.Spec.Containers[0].Env) {
runnerBusy, err := r.isRunnerBusy(ctx, runner.Spec.Organization, runner.Spec.Repository, runner.Name)
if err != nil {
log.Error(err, "Failed to check if runner is busy")
return ctrl.Result{}, nil
}
if !runnerBusy && (!reflect.DeepEqual(pod.Spec.Containers[0].Env, newPod.Spec.Containers[0].Env) || pod.Spec.Containers[0].Image != newPod.Spec.Containers[0].Image) {
restart = true
}
if !restart {
return ctrl.Result{}, err
}
@@ -225,6 +229,21 @@ func (r *RunnerReconciler) Reconcile(req ctrl.Request) (ctrl.Result, error) {
return ctrl.Result{}, nil
}
func (r *RunnerReconciler) isRunnerBusy(ctx context.Context, org, repo, name string) (bool, error) {
runners, err := r.GitHubClient.ListRunners(ctx, org, repo)
if err != nil {
return false, err
}
for _, runner := range runners {
if runner.GetName() == name {
return runner.GetBusy(), nil
}
}
return false, fmt.Errorf("runner not found")
}
func (r *RunnerReconciler) unregisterRunner(ctx context.Context, org, repo, name string) (bool, error) {
runners, err := r.GitHubClient.ListRunners(ctx, org, repo)
if err != nil {
@@ -234,6 +253,9 @@ func (r *RunnerReconciler) unregisterRunner(ctx context.Context, org, repo, name
id := int64(0)
for _, runner := range runners {
if runner.GetName() == name {
if runner.GetBusy() {
return false, fmt.Errorf("runner is busy")
}
id = runner.GetID()
break
}

View File

@@ -31,14 +31,16 @@ import (
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"github.com/summerwind/actions-runner-controller/api/v1alpha1"
"github.com/summerwind/actions-runner-controller/github"
)
// RunnerReplicaSetReconciler reconciles a Runner object
type RunnerReplicaSetReconciler struct {
client.Client
Log logr.Logger
Recorder record.EventRecorder
Scheme *runtime.Scheme
Log logr.Logger
Recorder record.EventRecorder
Scheme *runtime.Scheme
GitHubClient *github.Client
}
// +kubebuilder:rbac:groups=actions.summerwind.dev,resources=runnerreplicasets,verbs=get;list;watch;create;update;patch;delete
@@ -96,8 +98,25 @@ func (r *RunnerReplicaSetReconciler) Reconcile(req ctrl.Request) (ctrl.Result, e
if available > desired {
n := available - desired
// get runners that are currently not busy
var notBusy []v1alpha1.Runner
for _, runner := range myRunners {
busy, err := r.isRunnerBusy(ctx, runner.Spec.Organization, runner.Spec.Repository, runner.Name)
if err != nil {
log.Error(err, "Failed to check if runner is busy")
return ctrl.Result{}, err
}
if !busy {
notBusy = append(notBusy, runner)
}
}
if len(notBusy) < n {
n = len(notBusy)
}
for i := 0; i < n; i++ {
if err := r.Client.Delete(ctx, &myRunners[i]); err != nil {
if err := r.Client.Delete(ctx, &notBusy[i]); err != nil {
log.Error(err, "Failed to delete runner resource")
return ctrl.Result{}, err
@@ -166,3 +185,19 @@ func (r *RunnerReplicaSetReconciler) SetupWithManager(mgr ctrl.Manager) error {
Owns(&v1alpha1.Runner{}).
Complete(r)
}
func (r *RunnerReplicaSetReconciler) isRunnerBusy(ctx context.Context, org, repo, name string) (bool, error) {
runners, err := r.GitHubClient.ListRunners(ctx, org, repo)
r.Log.Info("runners", "github", runners)
if err != nil {
return false, err
}
for _, runner := range runners {
if runner.GetName() == name {
return runner.GetBusy(), nil
}
}
return false, fmt.Errorf("runner not found")
}

View File

@@ -3,11 +3,14 @@ package controllers
import (
"context"
"math/rand"
"net/http/httptest"
"time"
"github.com/google/go-github/v32/github"
corev1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/types"
"k8s.io/client-go/kubernetes/scheme"
"k8s.io/utils/pointer"
ctrl "sigs.k8s.io/controller-runtime"
logf "sigs.k8s.io/controller-runtime/pkg/log"
@@ -17,6 +20,12 @@ import (
"sigs.k8s.io/controller-runtime/pkg/client"
actionsv1alpha1 "github.com/summerwind/actions-runner-controller/api/v1alpha1"
"github.com/summerwind/actions-runner-controller/github/fake"
)
var (
runnersList *fake.RunnersList
server *httptest.Server
)
// SetupTest will set up a testing environment.
@@ -41,11 +50,16 @@ func SetupTest(ctx context.Context) *corev1.Namespace {
mgr, err := ctrl.NewManager(cfg, ctrl.Options{})
Expect(err).NotTo(HaveOccurred(), "failed to create manager")
runnersList = fake.NewRunnersList()
server = runnersList.GetServer()
ghClient := newGithubClient(server)
controller := &RunnerReplicaSetReconciler{
Client: mgr.GetClient(),
Scheme: scheme.Scheme,
Log: logf.Log,
Recorder: mgr.GetEventRecorderFor("runnerreplicaset-controller"),
Client: mgr.GetClient(),
Scheme: scheme.Scheme,
Log: logf.Log,
Recorder: mgr.GetEventRecorderFor("runnerreplicaset-controller"),
GitHubClient: ghClient,
}
err = controller.SetupWithManager(mgr)
Expect(err).NotTo(HaveOccurred(), "failed to setup controller")
@@ -61,6 +75,7 @@ func SetupTest(ctx context.Context) *corev1.Namespace {
AfterEach(func() {
close(stopCh)
server.Close()
err := k8sClient.Delete(ctx, ns)
Expect(err).NotTo(HaveOccurred(), "failed to delete test namespace")
})
@@ -124,6 +139,16 @@ var _ = Context("Inside of a new namespace", func() {
logf.Log.Error(err, "list runners")
}
for i, runner := range runners.Items {
runnersList.Add(&github.Runner{
ID: pointer.Int64Ptr(int64(i) + 1),
Name: pointer.StringPtr(runner.Name),
OS: pointer.StringPtr("linux"),
Status: pointer.StringPtr("online"),
Busy: pointer.BoolPtr(false),
})
}
return len(runners.Items)
},
time.Second*5, time.Millisecond*500).Should(BeEquivalentTo(1))
@@ -155,6 +180,16 @@ var _ = Context("Inside of a new namespace", func() {
logf.Log.Error(err, "list runners")
}
for i, runner := range runners.Items {
runnersList.Add(&github.Runner{
ID: pointer.Int64Ptr(int64(i) + 1),
Name: pointer.StringPtr(runner.Name),
OS: pointer.StringPtr("linux"),
Status: pointer.StringPtr("online"),
Busy: pointer.BoolPtr(false),
})
}
return len(runners.Items)
},
time.Second*5, time.Millisecond*500).Should(BeEquivalentTo(2))
@@ -186,6 +221,16 @@ var _ = Context("Inside of a new namespace", func() {
logf.Log.Error(err, "list runners")
}
for i, runner := range runners.Items {
runnersList.Add(&github.Runner{
ID: pointer.Int64Ptr(int64(i) + 1),
Name: pointer.StringPtr(runner.Name),
OS: pointer.StringPtr("linux"),
Status: pointer.StringPtr("online"),
Busy: pointer.BoolPtr(false),
})
}
return len(runners.Items)
},
time.Second*5, time.Millisecond*500).Should(BeEquivalentTo(0))