Compare commits

...

6 Commits

Author SHA1 Message Date
Yusuke Kuoka
434823bcb3 scale{Up,Down}Adjustment to add/remove constant number of replicas on scaling (#315)
* `scale{Up,Down}Adjustment` to add/remove constant number of replicas on scaling

Ref #305

* Bump chart version
2021-02-16 17:16:26 +09:00
Yusuke Kuoka
35d047db01 Fix enterprise runners misusing cached token (#314)
Follow-up for #290
2021-02-16 12:56:52 +09:00
Yusuke Kuoka
f1db6af1c5 Add repository runners support for PercentageRunnersBusy-based autoscaling (#313)
Resolves #258
2021-02-16 12:44:51 +09:00
Hidetake Iwata
4f3f2fb60d Add metrics for GitHub API rate limit (#312) 2021-02-16 09:58:09 +09:00
Johannes Nicolai
2623140c9a Make log message less scary (#311)
* the reconciliation loop is often much faster than the runner startup, 
so changing runner not found related messages to debug and also add the 
possibility that the runner just needs more time
2021-02-16 09:55:55 +09:00
Johannes Nicolai
1db9d9d574 Use ARM64 compatible kube-rbac-proxy from upstream (#310)
* as pointed out in #281 the currently used image for the 
kube-rbac-proxy - gcr.io/kubebuilder/kube-rbac-proxy:v0.4.1" - does not 
have an ARM64 image
* hence, trying to use the standard deployment manifest / helm char will 
fail on ARM64 systems
* replaced image with quay.io/brancz/kube-rbac-proxy:v0.8.0 which is the 
latest version from the upstream maintainer 
(https://github.com/brancz/kube-rbac-proxy/blob/master/Makefile#L13)
* successfully tested on both AMD64 and ARM64 clusters
* fixes #281
2021-02-16 09:55:03 +09:00
13 changed files with 198 additions and 27 deletions

View File

@@ -126,6 +126,16 @@ type MetricSpec struct {
// to determine how many pods should be removed.
// +optional
ScaleDownFactor string `json:"scaleDownFactor,omitempty"`
// ScaleUpAdjustment is the number of runners added on scale-up.
// You can only specify either ScaleUpFactor or ScaleUpAdjustment.
// +optional
ScaleUpAdjustment int `json:"scaleUpAdjustment,omitempty"`
// ScaleDownAdjustment is the number of runners removed on scale-down.
// You can only specify either ScaleDownFactor or ScaleDownAdjustment.
// +optional
ScaleDownAdjustment int `json:"scaleDownAdjustment,omitempty"`
}
type HorizontalRunnerAutoscalerStatus struct {

View File

@@ -15,7 +15,7 @@ type: application
# This is the chart version. This version number should be incremented each time you make changes
# to the chart and its templates, including the app version.
# Versions are expected to follow Semantic Versioning (https://semver.org/)
version: 0.4.0
version: 0.5.0
home: https://github.com/summerwind/actions-runner-controller

View File

@@ -78,6 +78,11 @@ spec:
items:
type: string
type: array
scaleDownAdjustment:
description: ScaleDownAdjustment is the number of runners removed
on scale-down. You can only specify either ScaleDownFactor or
ScaleDownAdjustment.
type: integer
scaleDownFactor:
description: ScaleDownFactor is the multiplicative factor applied
to the current number of runners used to determine how many
@@ -87,6 +92,10 @@ spec:
description: ScaleDownThreshold is the percentage of busy runners
less than which will trigger the hpa to scale the runners down.
type: string
scaleUpAdjustment:
description: ScaleUpAdjustment is the number of runners added
on scale-up. You can only specify either ScaleUpFactor or ScaleUpAdjustment.
type: integer
scaleUpFactor:
description: ScaleUpFactor is the multiplicative factor applied
to the current number of runners used to determine how many

View File

@@ -28,8 +28,8 @@ image:
kube_rbac_proxy:
image:
repository: gcr.io/kubebuilder/kube-rbac-proxy
tag: v0.4.1
repository: quay.io/brancz/kube-rbac-proxy
tag: v0.8.0
imagePullSecrets: []
nameOverride: ""

View File

@@ -78,6 +78,11 @@ spec:
items:
type: string
type: array
scaleDownAdjustment:
description: ScaleDownAdjustment is the number of runners removed
on scale-down. You can only specify either ScaleDownFactor or
ScaleDownAdjustment.
type: integer
scaleDownFactor:
description: ScaleDownFactor is the multiplicative factor applied
to the current number of runners used to determine how many
@@ -87,6 +92,10 @@ spec:
description: ScaleDownThreshold is the percentage of busy runners
less than which will trigger the hpa to scale the runners down.
type: string
scaleUpAdjustment:
description: ScaleUpAdjustment is the number of runners added
on scale-up. You can only specify either ScaleUpFactor or ScaleUpAdjustment.
type: integer
scaleUpFactor:
description: ScaleUpFactor is the multiplicative factor applied
to the current number of runners used to determine how many

View File

@@ -10,7 +10,7 @@ spec:
spec:
containers:
- name: kube-rbac-proxy
image: gcr.io/kubebuilder/kube-rbac-proxy:v0.4.1
image: quay.io/brancz/kube-rbac-proxy:v0.8.0
args:
- "--secure-listen-address=0.0.0.0:8443"
- "--upstream=http://127.0.0.1:8080/"

View File

@@ -189,6 +189,9 @@ func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByQueuedAndInPro
"workflow_runs_in_progress", inProgress,
"workflow_runs_queued", queued,
"workflow_runs_unknown", unknown,
"namespace", hra.Namespace,
"runner_deployment", rd.Name,
"horizontal_runner_autoscaler", hra.Name,
)
return &replicas, nil
@@ -196,7 +199,6 @@ func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByQueuedAndInPro
func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByPercentageRunnersBusy(rd v1alpha1.RunnerDeployment, hra v1alpha1.HorizontalRunnerAutoscaler) (*int, error) {
ctx := context.Background()
orgName := rd.Spec.Template.Spec.Organization
minReplicas := *hra.Spec.MinReplicas
maxReplicas := *hra.Spec.MaxReplicas
metrics := hra.Spec.Metrics[0]
@@ -220,14 +222,34 @@ func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByPercentageRunn
scaleDownThreshold = sdt
}
if metrics.ScaleUpFactor != "" {
scaleUpAdjustment := metrics.ScaleUpAdjustment
if scaleUpAdjustment != 0 {
if metrics.ScaleUpAdjustment < 0 {
return nil, errors.New("validating autoscaling metrics: spec.autoscaling.metrics[].scaleUpAdjustment cannot be lower than 0")
}
if metrics.ScaleUpFactor != "" {
return nil, errors.New("validating autoscaling metrics: spec.autoscaling.metrics[]: scaleUpAdjustment and scaleUpFactor cannot be specified together")
}
} else if metrics.ScaleUpFactor != "" {
suf, err := strconv.ParseFloat(metrics.ScaleUpFactor, 64)
if err != nil {
return nil, errors.New("validating autoscaling metrics: spec.autoscaling.metrics[].scaleUpFactor cannot be parsed into a float64")
}
scaleUpFactor = suf
}
if metrics.ScaleDownFactor != "" {
scaleDownAdjustment := metrics.ScaleDownAdjustment
if scaleDownAdjustment != 0 {
if metrics.ScaleDownAdjustment < 0 {
return nil, errors.New("validating autoscaling metrics: spec.autoscaling.metrics[].scaleDownAdjustment cannot be lower than 0")
}
if metrics.ScaleDownFactor != "" {
return nil, errors.New("validating autoscaling metrics: spec.autoscaling.metrics[]: scaleDownAdjustment and scaleDownFactor cannot be specified together")
}
} else if metrics.ScaleDownFactor != "" {
sdf, err := strconv.ParseFloat(metrics.ScaleDownFactor, 64)
if err != nil {
return nil, errors.New("validating autoscaling metrics: spec.autoscaling.metrics[].scaleDownFactor cannot be parsed into a float64")
@@ -245,8 +267,18 @@ func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByPercentageRunn
runnerMap[items.Name] = struct{}{}
}
var (
enterprise = rd.Spec.Template.Spec.Enterprise
organization = rd.Spec.Template.Spec.Organization
repository = rd.Spec.Template.Spec.Repository
)
// ListRunners will return all runners managed by GitHub - not restricted to ns
runners, err := r.GitHubClient.ListRunners(ctx, "", orgName, "")
runners, err := r.GitHubClient.ListRunners(
ctx,
enterprise,
organization,
repository)
if err != nil {
return nil, err
}
@@ -261,9 +293,17 @@ func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByPercentageRunn
var desiredReplicas int
fractionBusy := float64(numRunnersBusy) / float64(numRunners)
if fractionBusy >= scaleUpThreshold {
desiredReplicas = int(math.Ceil(float64(numRunners) * scaleUpFactor))
if scaleUpAdjustment > 0 {
desiredReplicas = numRunners + scaleUpAdjustment
} else {
desiredReplicas = int(math.Ceil(float64(numRunners) * scaleUpFactor))
}
} else if fractionBusy < scaleDownThreshold {
desiredReplicas = int(float64(numRunners) * scaleDownFactor)
if scaleDownAdjustment > 0 {
desiredReplicas = numRunners - scaleDownAdjustment
} else {
desiredReplicas = int(float64(numRunners) * scaleDownFactor)
}
} else {
desiredReplicas = *rd.Spec.Replicas
}
@@ -282,6 +322,12 @@ func (r *HorizontalRunnerAutoscalerReconciler) calculateReplicasByPercentageRunn
"current_replicas", rd.Spec.Replicas,
"num_runners", numRunners,
"num_runners_busy", numRunnersBusy,
"namespace", hra.Namespace,
"runner_deployment", rd.Name,
"horizontal_runner_autoscaler", hra.Name,
"enterprise", enterprise,
"organization", organization,
"repository", repository,
)
rd.Status.Replicas = &desiredReplicas

View File

@@ -250,7 +250,7 @@ func (r *RunnerReconciler) Reconcile(req ctrl.Request) (ctrl.Result, error) {
if err != nil {
var e *github.RunnerNotFound
if errors.As(err, &e) {
log.Error(err, "Failed to check if runner is busy. Probably this runner has never been successfully registered to GitHub.")
log.V(1).Info("Failed to check if runner is busy. Either this runner has never been successfully registered to GitHub or it still needs more time.", "runnerName", runner.Name)
notRegistered = true
} else {

View File

@@ -111,7 +111,7 @@ func (r *RunnerReplicaSetReconciler) Reconcile(req ctrl.Request) (ctrl.Result, e
var e *github.RunnerNotFound
if errors.As(err, &e) {
log.Error(err, "Failed to check if runner is busy. Probably this runner has never been successfully registered to GitHub, and therefore we prioritize it for deletion", "runnerName", runner.Name)
log.V(1).Info("Failed to check if runner is busy. Either this runner has never been successfully registered to GitHub or has not managed yet to, and therefore we prioritize it for deletion", "runnerName", runner.Name)
notRegistered = true
} else {
var e *gogithub.RateLimitError

View File

@@ -102,6 +102,18 @@ func NewServer(opts ...Option) *httptest.Server {
Status: http.StatusBadRequest,
Body: "",
},
"/enterprises/test/actions/runners/registration-token": &Handler{
Status: http.StatusCreated,
Body: fmt.Sprintf("{\"token\": \"%s\", \"expires_at\": \"%s\"}", RegistrationToken, time.Now().Add(time.Hour*1).Format(time.RFC3339)),
},
"/enterprises/invalid/actions/runners/registration-token": &Handler{
Status: http.StatusOK,
Body: fmt.Sprintf("{\"token\": \"%s\", \"expires_at\": \"%s\"}", RegistrationToken, time.Now().Add(time.Hour*1).Format(time.RFC3339)),
},
"/enterprises/error/actions/runners/registration-token": &Handler{
Status: http.StatusBadRequest,
Body: "",
},
// For ListRunners
"/repos/test/valid/actions/runners": config.FixedResponses.ListRunners,
@@ -125,6 +137,18 @@ func NewServer(opts ...Option) *httptest.Server {
Status: http.StatusBadRequest,
Body: "",
},
"/enterprises/test/actions/runners": &Handler{
Status: http.StatusOK,
Body: RunnersListBody,
},
"/enterprises/invalid/actions/runners": &Handler{
Status: http.StatusNoContent,
Body: "",
},
"/enterprises/error/actions/runners": &Handler{
Status: http.StatusBadRequest,
Body: "",
},
// For RemoveRunner
"/repos/test/valid/actions/runners/1": &Handler{
@@ -151,6 +175,18 @@ func NewServer(opts ...Option) *httptest.Server {
Status: http.StatusBadRequest,
Body: "",
},
"/enterprises/test/actions/runners/1": &Handler{
Status: http.StatusNoContent,
Body: "",
},
"/enterprises/invalid/actions/runners/1": &Handler{
Status: http.StatusOK,
Body: "",
},
"/enterprises/error/actions/runners/1": &Handler{
Status: http.StatusBadRequest,
Body: "",
},
// For auto-scaling based on the number of queued(pending) workflow runs
"/repos/test/valid/actions/runs": config.FixedResponses.ListRepositoryWorkflowRuns,

View File

@@ -11,6 +11,7 @@ import (
"github.com/bradleyfalzon/ghinstallation"
"github.com/google/go-github/v33/github"
"github.com/summerwind/actions-runner-controller/github/metrics"
"golang.org/x/oauth2"
)
@@ -34,15 +35,9 @@ type Client struct {
// NewClient creates a Github Client
func (c *Config) NewClient() (*Client, error) {
var (
httpClient *http.Client
client *github.Client
)
githubBaseURL := "https://github.com/"
var transport http.RoundTripper
if len(c.Token) > 0 {
httpClient = oauth2.NewClient(context.Background(), oauth2.StaticTokenSource(
&oauth2.Token{AccessToken: c.Token},
))
transport = oauth2.NewClient(context.Background(), oauth2.StaticTokenSource(&oauth2.Token{AccessToken: c.Token})).Transport
} else {
tr, err := ghinstallation.NewKeyFromFile(http.DefaultTransport, c.AppID, c.AppInstallationID, c.AppPrivateKey)
if err != nil {
@@ -55,9 +50,13 @@ func (c *Config) NewClient() (*Client, error) {
}
tr.BaseURL = githubAPIURL
}
httpClient = &http.Client{Transport: tr}
transport = tr
}
transport = metrics.Transport{Transport: transport}
httpClient := &http.Client{Transport: transport}
var client *github.Client
var githubBaseURL string
if len(c.EnterpriseURL) > 0 {
var err error
client, err = github.NewEnterpriseClient(c.EnterpriseURL, c.EnterpriseURL, httpClient)
@@ -67,6 +66,7 @@ func (c *Config) NewClient() (*Client, error) {
githubBaseURL = fmt.Sprintf("%s://%s%s", client.BaseURL.Scheme, client.BaseURL.Host, strings.TrimSuffix(client.BaseURL.Path, "api/v3/"))
} else {
client = github.NewClient(httpClient)
githubBaseURL = "https://github.com/"
}
return &Client{
@@ -82,7 +82,7 @@ func (c *Client) GetRegistrationToken(ctx context.Context, enterprise, org, repo
c.mu.Lock()
defer c.mu.Unlock()
key := getRegistrationKey(org, repo)
key := getRegistrationKey(org, repo, enterprise)
rt, ok := c.regTokens[key]
if ok && rt.GetExpiresAt().After(time.Now()) {
@@ -250,11 +250,8 @@ func getEnterpriseOrganisationAndRepo(enterprise, org, repo string) (string, str
return "", "", "", fmt.Errorf("enterprise, organization and repository are all empty")
}
func getRegistrationKey(org, repo string) string {
if len(org) > 0 {
return org
}
return repo
func getRegistrationKey(org, repo, enterprise string) string {
return fmt.Sprintf("org=%s,repo=%s,enterprise=%s", org, repo, enterprise)
}
func splitOwnerAndRepo(repo string) (string, string, error) {

View File

@@ -0,0 +1,63 @@
// Package metrics provides monitoring of the GitHub related metrics.
//
// This depends on the metrics exporter of kubebuilder.
// See https://book.kubebuilder.io/reference/metrics.html for details.
package metrics
import (
"net/http"
"strconv"
"github.com/prometheus/client_golang/prometheus"
"sigs.k8s.io/controller-runtime/pkg/metrics"
)
func init() {
metrics.Registry.MustRegister(metricRateLimit, metricRateLimitRemaining)
}
var (
// https://docs.github.com/en/rest/overview/resources-in-the-rest-api#rate-limiting
metricRateLimit = prometheus.NewGauge(
prometheus.GaugeOpts{
Name: "github_rate_limit",
Help: "The maximum number of requests you're permitted to make per hour",
},
)
metricRateLimitRemaining = prometheus.NewGauge(
prometheus.GaugeOpts{
Name: "github_rate_limit_remaining",
Help: "The number of requests remaining in the current rate limit window",
},
)
)
const (
// https://docs.github.com/en/rest/overview/resources-in-the-rest-api#rate-limiting
headerRateLimit = "X-RateLimit-Limit"
headerRateLimitRemaining = "X-RateLimit-Remaining"
)
// Transport wraps a transport with metrics monitoring
type Transport struct {
Transport http.RoundTripper
}
func (t Transport) RoundTrip(req *http.Request) (*http.Response, error) {
resp, err := t.Transport.RoundTrip(req)
if resp != nil {
parseResponse(resp)
}
return resp, err
}
func parseResponse(resp *http.Response) {
rateLimit, err := strconv.Atoi(resp.Header.Get(headerRateLimit))
if err == nil {
metricRateLimit.Set(float64(rateLimit))
}
rateLimitRemaining, err := strconv.Atoi(resp.Header.Get(headerRateLimitRemaining))
if err == nil {
metricRateLimitRemaining.Set(float64(rateLimitRemaining))
}
}

1
go.mod
View File

@@ -11,6 +11,7 @@ require (
github.com/kelseyhightower/envconfig v1.4.0
github.com/onsi/ginkgo v1.8.0
github.com/onsi/gomega v1.5.0
github.com/prometheus/client_golang v0.9.2
github.com/stretchr/testify v1.4.0 // indirect
golang.org/x/oauth2 v0.0.0-20190604053449-0f29369cfe45
k8s.io/api v0.0.0-20190918155943-95b840bb6a1f