Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Use keep-alive for direct pod scraping #8367

Merged
merged 1 commit into from
Jun 18, 2020
Merged

Use keep-alive for direct pod scraping #8367

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 26 additions & 14 deletions pkg/autoscaler/metrics/stats_scraper.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,22 +102,32 @@ type scrapeClient interface {
Scrape(url string) (Stat, error)
}

// cacheDisabledClient is a http client with cache disabled. It is shared by
// every goruntime for a revision scraper.
var cacheDisabledClient = &http.Client{
// noKeepaliveClient is a http client with HTTP Keep-Alive disabled.
// This client is used in the mesh case since we want to get a new connection -
// and therefore, hopefully, host - on every scrape of the service.
var noKeepaliveClient = &http.Client{
Transport: &http.Transport{
// Do not use the cached connection
DisableKeepAlives: true,
},
Timeout: httpClientTimeout,
}

// client is a normal http client with HTTP Keep-Alive enabled.
// This client is used in the direct pod scraping (no mesh) case where we want
// to take advantage of HTTP Keep-Alive to avoid connection creation overhead
// between scrapes of the same pod.
var client = &http.Client{
Timeout: httpClientTimeout,
}

// serviceScraper scrapes Revision metrics via a K8S service by sampling. Which
// pod to be picked up to serve the request is decided by K8S. Please see
// https://kubernetes.io/docs/concepts/services-networking/network-policies/
// for details.
type serviceScraper struct {
sClient scrapeClient
directClient scrapeClient
meshClient scrapeClient

counter resources.EndpointsCounter
url string
statsCtx context.Context
Expand All @@ -131,28 +141,29 @@ type serviceScraper struct {
// the given Metric is responsible for.
func NewStatsScraper(metric *av1alpha1.Metric, counter resources.EndpointsCounter,
podAccessor resources.PodAccessor, logger *zap.SugaredLogger) (StatsScraper, error) {
sClient, err := newHTTPScrapeClient(cacheDisabledClient)
directClient, err := newHTTPScrapeClient(client)
if err != nil {
return nil, err
}
return newServiceScraperWithClient(metric, counter, podAccessor, sClient, logger)
meshClient, err := newHTTPScrapeClient(noKeepaliveClient)
if err != nil {
return nil, err
}
return newServiceScraperWithClient(metric, counter, podAccessor, directClient, meshClient, logger)
}

func newServiceScraperWithClient(
metric *av1alpha1.Metric,
counter resources.EndpointsCounter,
podAccessor resources.PodAccessor,
sClient scrapeClient,
directClient, meshClient scrapeClient,
logger *zap.SugaredLogger) (*serviceScraper, error) {
if metric == nil {
return nil, errors.New("metric must not be nil")
}
if counter == nil {
return nil, errors.New("counter must not be nil")
}
if sClient == nil {
return nil, errors.New("scrape client must not be nil")
}
revName := metric.Labels[serving.RevisionLabelKey]
if revName == "" {
return nil, errors.New("no Revision label found for Metric " + metric.Name)
Expand All @@ -166,7 +177,8 @@ func newServiceScraperWithClient(
}

return &serviceScraper{
sClient: sClient,
directClient: directClient,
meshClient: meshClient,
counter: counter,
url: urlFromTarget(metric.Spec.ScrapeTarget, metric.ObjectMeta.Namespace),
podAccessor: podAccessor,
Expand Down Expand Up @@ -253,7 +265,7 @@ func (s *serviceScraper) scrapePods(readyPods int) (Stat, error) {

// Scrape!
target := "http://" + pods[myIdx] + ":" + portAndPath
stat, err := s.sClient.Scrape(target)
stat, err := s.directClient.Scrape(target)
if err == nil {
results <- stat
return nil
Expand Down Expand Up @@ -388,7 +400,7 @@ func (s *serviceScraper) scrapeService(window time.Duration, readyPods int) (Sta
// tryScrape runs a single scrape and returns stat if this is a pod that has not been
// seen before. An error otherwise or if scraping failed.
func (s *serviceScraper) tryScrape(scrapedPods *sync.Map) (Stat, error) {
stat, err := s.sClient.Scrape(s.url)
stat, err := s.meshClient.Scrape(s.url)
if err != nil {
return emptyStat, err
}
Expand Down
51 changes: 25 additions & 26 deletions pkg/autoscaler/metrics/stats_scraper_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,12 +152,6 @@ func TestNewServiceScraperWithClientErrorCases(t *testing.T) {
counter: counter,
accessor: podAccessor,
expectedErr: "no Revision label found for Metric test-revision",
}, {
name: "Empty scrape client",
metric: metric,
counter: counter,
accessor: podAccessor,
expectedErr: "scrape client must not be nil",
}, {
name: "Empty counter",
metric: metric,
Expand All @@ -170,7 +164,7 @@ func TestNewServiceScraperWithClientErrorCases(t *testing.T) {
for _, test := range testCases {
t.Run(test.name, func(t *testing.T) {
if _, err := newServiceScraperWithClient(test.metric, test.counter,
test.accessor, test.client, logger); err != nil {
test.accessor, test.client, test.client, logger); err != nil {
got := err.Error()
want := test.expectedErr
if got != want {
Expand All @@ -189,7 +183,7 @@ func TestPodDirectScrapeSuccess(t *testing.T) {
fake.KubeInformer = kubeinformers.NewSharedInformerFactory(fake.KubeClient, 0)

client := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, client, true)
scraper, err := serviceScraperForTest(t, client, nil /* mesh not used */, true)
if err != nil {
t.Fatal("serviceScraperForTest:", err)
}
Expand All @@ -216,7 +210,7 @@ func TestPodDirectScrapeSomeFailButSuccess(t *testing.T) {

// For 5 pods, we need 4 successes.
client := newTestScrapeClient(testStats, []error{nil, nil, errors.New("okay"), nil, nil})
scraper, err := serviceScraperForTest(t, client, true)
scraper, err := serviceScraperForTest(t, client, nil /* mesh not used */, true)
if err != nil {
t.Fatal("serviceScraperForTest:", err)
}
Expand Down Expand Up @@ -248,13 +242,15 @@ func TestPodDirectScrapeNoneSucceed(t *testing.T) {
fake.KubeInformer = kubeinformers.NewSharedInformerFactory(fake.KubeClient, 0)

testStats := testStatsWithTime(4, youngPodCutOffDuration.Seconds() /*youngest*/)
client := newTestScrapeClient(testStats, []error{
direct := newTestScrapeClient(testStats, []error{
// Pods fail.
errors.New("okay"), errors.New("okay"), errors.New("okay"), errors.New("okay"),
})
mesh := newTestScrapeClient(testStats, []error{
// Service succeeds.
nil, nil, nil, nil,
})
scraper, err := serviceScraperForTest(t, client, true)
scraper, err := serviceScraperForTest(t, direct, mesh, true)
if err != nil {
t.Fatal("serviceScraperForTest:", err)
}
Expand Down Expand Up @@ -286,7 +282,7 @@ func TestPodDirectScrapePodsExhausted(t *testing.T) {
fake.KubeInformer = kubeinformers.NewSharedInformerFactory(fake.KubeClient, 0)

client := newTestScrapeClient(testStats, []error{nil, nil, errors.New("okay"), nil})
scraper, err := serviceScraperForTest(t, client, true)
scraper, err := serviceScraperForTest(t, client, nil /* mesh not used */, true)
if err != nil {
t.Fatal("serviceScraperForTest:", err)
}
Expand All @@ -305,7 +301,7 @@ func TestPodDirectScrapePodsExhausted(t *testing.T) {

func TestScrapeReportStatWhenAllCallsSucceed(t *testing.T) {
client := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, client, false)
scraper, err := serviceScraperForTest(t, client, client, false)
if err != nil {
t.Fatal("serviceScraperForTest:", err)
}
Expand Down Expand Up @@ -336,8 +332,9 @@ func TestScrapeAllPodsYoungPods(t *testing.T) {
// acceptable.
testStats := testStatsWithTime(numP, 0. /*youngest*/)

client := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, client, false)
direct := newTestScrapeClient(testStats, []error{errNoPodsScraped}) // fall back to service scrape
mesh := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, direct, mesh, false)
if err != nil {
t.Fatalf("serviceScraperForTest=%v, want no error", err)
}
Expand Down Expand Up @@ -368,8 +365,9 @@ func TestScrapeAllPodsOldPods(t *testing.T) {
// All pods are at least cutoff time old, so first 5 stats will be picked.
testStats := testStatsWithTime(numP, youngPodCutOffDuration.Seconds() /*youngest*/)

client := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, client, false)
direct := newTestScrapeClient(testStats, []error{errNoPodsScraped}) // fall back to service scrape
mesh := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, direct, mesh, false)
if err != nil {
t.Fatalf("serviceScraperForTest=%v, want no error", err)
}
Expand Down Expand Up @@ -401,8 +399,9 @@ func TestScrapeSomePodsOldPods(t *testing.T) {
// So pods 3-10 qualify (for 11 total sample is 7).
testStats := testStatsWithTime(numP, youngPodCutOffDuration.Seconds()/2 /*youngest*/)

client := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, client, false)
direct := newTestScrapeClient(testStats, []error{errNoPodsScraped}) // fall back to service scrape
mesh := newTestScrapeClient(testStats, []error{nil})
scraper, err := serviceScraperForTest(t, direct, mesh, false)
if err != nil {
t.Fatalf("serviceScraperForTest=%v, want no error", err)
}
Expand Down Expand Up @@ -430,7 +429,7 @@ func TestScrapeSomePodsOldPods(t *testing.T) {

func TestScrapeReportErrorCannotFindEnoughPods(t *testing.T) {
client := newTestScrapeClient(testStats[2:], []error{nil})
scraper, err := serviceScraperForTest(t, client, false)
scraper, err := serviceScraperForTest(t, client, client, false)
if err != nil {
t.Fatalf("serviceScraperForTest=%v, want no error", err)
}
Expand All @@ -448,9 +447,9 @@ func TestScrapeReportErrorIfAnyFails(t *testing.T) {
errTest := errors.New("test")

// 1 success and 10 failures so one scrape fails permanently through retries.
client := newTestScrapeClient(testStats, []error{nil,
errTest, errTest, errTest, errTest, errTest, errTest, errTest, errTest, errTest, errTest})
scraper, err := serviceScraperForTest(t, client, false)
client := newTestScrapeClient(testStats, []error{nil, errTest, errTest,
errTest, errTest, errTest, errTest, errTest, errTest, errTest, errTest})
scraper, err := serviceScraperForTest(t, client, client, false)
if err != nil {
t.Fatalf("serviceScraperForTest=%v, want no error", err)
}
Expand All @@ -466,7 +465,7 @@ func TestScrapeReportErrorIfAnyFails(t *testing.T) {

func TestScrapeDoNotScrapeIfNoPodsFound(t *testing.T) {
client := newTestScrapeClient(testStats, nil)
scraper, err := serviceScraperForTest(t, client, false)
scraper, err := serviceScraperForTest(t, client, client, false)
if err != nil {
t.Fatalf("serviceScraperForTest=%v, want no error", err)
}
Expand All @@ -483,7 +482,7 @@ func TestScrapeDoNotScrapeIfNoPodsFound(t *testing.T) {
}
}

func serviceScraperForTest(t *testing.T, sClient scrapeClient, podsAddressable bool) (*serviceScraper, error) {
func serviceScraperForTest(t *testing.T, directClient, meshClient scrapeClient, podsAddressable bool) (*serviceScraper, error) {
metric := testMetric()
counter := resources.NewScopedEndpointsCounter(
fake.KubeInformer.Core().V1().Endpoints().Lister(),
Expand All @@ -492,7 +491,7 @@ func serviceScraperForTest(t *testing.T, sClient scrapeClient, podsAddressable b
fake.KubeInformer.Core().V1().Pods().Lister(),
fake.TestNamespace, fake.TestRevision)
logger := logtesting.TestLogger(t)
ss, err := newServiceScraperWithClient(metric, counter, accessor, sClient, logger)
ss, err := newServiceScraperWithClient(metric, counter, accessor, directClient, meshClient, logger)
if ss != nil {
ss.podsAddressable = podsAddressable
}
Expand Down