Added pods-per-core to kubelet. #25762

cmd/integration/integration.go

@@ -240,6 +240,7 @@ func startComponents(firstManifestURL, secondManifestURL string) (string, string
 		10*time.Second, /* OutOfDiskTransitionFrequency */
 		10*time.Second, /* EvictionPressureTransitionPeriod */
 		40,             /* MaxPods */
+		0,              /* PodsPerCore*/
 		cm, net.ParseIP("127.0.0.1"))
 
 	kubeletapp.RunKubelet(kcfg)
@@ -273,6 +274,7 @@ func startComponents(firstManifestURL, secondManifestURL string) (string, string
 		10*time.Second, /* OutOfDiskTransitionFrequency */
 		10*time.Second, /* EvictionPressureTransitionPeriod */
 		40,             /* MaxPods */
+		0,              /* PodsPerCore*/
 		cm,
 		net.ParseIP("127.0.0.1"))
 

cmd/kubelet/app/options/options.go

@@ -146,6 +146,7 @@ func NewKubeletServer() *KubeletServer {
 			HairpinMode:                      componentconfig.PromiscuousBridge,
 			BabysitDaemons:                   false,
 			EvictionPressureTransitionPeriod: unversioned.Duration{Duration: 5 * time.Minute},
+			PodsPerCore:                      0,
 		},
 	}
 }
@@ -264,4 +265,5 @@ func (s *KubeletServer) AddFlags(fs *pflag.FlagSet) {
 	fs.StringVar(&s.EvictionSoftGracePeriod, "eviction-soft-grace-period", s.EvictionSoftGracePeriod, "A set of eviction grace periods (e.g. memory.available=1m30s) that correspond to how long a soft eviction threshold must hold before triggering a pod eviction.")
 	fs.DurationVar(&s.EvictionPressureTransitionPeriod.Duration, "eviction-pressure-transition-period", s.EvictionPressureTransitionPeriod.Duration, "Duration for which the kubelet has to wait before transitioning out of an eviction pressure condition.")
 	fs.Int32Var(&s.EvictionMaxPodGracePeriod, "eviction-max-pod-grace-period", s.EvictionMaxPodGracePeriod, "Maximum allowed grace period (in seconds) to use when terminating pods in response to a soft eviction threshold being met.  If negative, defer to pod specified value.")
+	fs.Int32Var(&s.PodsPerCore, "pods-per-core", s.PodsPerCore, "Number of Pods per core that can run on this Kubelet. The total number of Pods on this Kubelet cannot exceed max-pods, so max-pods will be used if this caulcation results in a larger number of Pods allowed on the Kubelet. A value of 0 disables this limit.")
 }

cmd/kubelet/app/server.go

@@ -277,6 +277,7 @@ func UnsecuredKubeletConfig(s *options.KubeletServer) (*KubeletConfig, error) {
 		ExperimentalFlannelOverlay:     s.ExperimentalFlannelOverlay,
 		NodeIP:         net.ParseIP(s.NodeIP),
 		EvictionConfig: evictionConfig,
+		PodsPerCore:    int(s.PodsPerCore),
 	}, nil
 }
 
@@ -533,7 +534,7 @@ func SimpleKubelet(client *clientset.Clientset,
 	cloud cloudprovider.Interface,
 	osInterface kubecontainer.OSInterface,
 	fileCheckFrequency, httpCheckFrequency, minimumGCAge, nodeStatusUpdateFrequency, syncFrequency, outOfDiskTransitionFrequency, evictionPressureTransitionPeriod time.Duration,
-	maxPods int,
+	maxPods int, podsPerCore int,
 	containerManager cm.ContainerManager, clusterDNS net.IP) *KubeletConfig {
 	imageGCPolicy := kubelet.ImageGCPolicy{
 		HighThresholdPercent: 90,
@@ -604,6 +605,7 @@ func SimpleKubelet(client *clientset.Clientset,
 		Writer:              &io.StdWriter{},
 		OutOfDiskTransitionFrequency: outOfDiskTransitionFrequency,
 		EvictionConfig:               evictionConfig,
+		PodsPerCore:                  podsPerCore,
 	}
 	return &kcfg
 }
@@ -814,6 +816,7 @@ type KubeletConfig struct {
 	OOMAdjuster                    *oom.OOMAdjuster
 	OSInterface                    kubecontainer.OSInterface
 	PodCIDR                        string
+	PodsPerCore                    int
 	ReconcileCIDR                  bool
 	PodConfig                      *config.PodConfig
 	PodInfraContainerImage         string
@@ -923,6 +926,7 @@ func CreateAndInitKubelet(kc *KubeletConfig) (k KubeletBootstrap, pc *config.Pod
 		kc.PodCIDR,
 		kc.ReconcileCIDR,
 		kc.MaxPods,
+		kc.PodsPerCore,
 		kc.NvidiaGPUs,
 		kc.DockerExecHandler,
 		kc.ResolverConfig,

cmd/kubemark/hollow-node.go

@@ -49,7 +49,8 @@ type HollowNodeConfig struct {
 }
 
 const (
-	maxPods = 110
+	maxPods     = 110
+	podsPerCore = 0
 )
 
 var knownMorphs = sets.NewString("kubelet", "proxy")
@@ -115,6 +116,7 @@ func main() {
 			config.KubeletReadOnlyPort,
 			containerManager,
 			maxPods,
+			podsPerCore,
 		)
 		hollowKubelet.Run()
 	}

docs/admin/kubelet.md

@@ -135,6 +135,7 @@ kubelet
       --outofdisk-transition-frequency=5m0s: Duration for which the kubelet has to wait before transitioning out of out-of-disk node condition status. Default: 5m0s
       --pod-cidr="": The CIDR to use for pod IP addresses, only used in standalone mode.  In cluster mode, this is obtained from the master.
       --pod-infra-container-image="gcr.io/google_containers/pause-amd64:3.0": The image whose network/ipc namespaces containers in each pod will use.
+      --pods-per-core=0: Number of Pods per core that can run on this Kubelet. The total number of Pods on this Kubelet cannot exceed max-pods, so max-pods will be used if this caulcation results in a larger number of Pods allowed on the Kubelet. A value of 0 disables this limit.
       --port=10250: The port for the Kubelet to serve on.
       --read-only-port=10255: The read-only port for the Kubelet to serve on with no authentication/authorization (set to 0 to disable)
       --really-crash-for-testing[=false]: If true, when panics occur crash. Intended for testing.

hack/verify-flags/known-flags.txt

@@ -329,6 +329,7 @@ pod-cidr
 pod-eviction-timeout
 pod-infra-container-image
 pod-running
+pods-per-core
 policy-config-file
 poll-interval
 portal-net

pkg/apis/componentconfig/deep_copy_generated.go

@@ -316,6 +316,7 @@ func DeepCopy_componentconfig_KubeletConfiguration(in KubeletConfiguration, out
 		return err
 	}
 	out.EvictionMaxPodGracePeriod = in.EvictionMaxPodGracePeriod
+	out.PodsPerCore = in.PodsPerCore
 	return nil
 }
 

pkg/apis/componentconfig/types.go

@@ -360,6 +360,8 @@ type KubeletConfiguration struct {
 	EvictionPressureTransitionPeriod unversioned.Duration `json:"evictionPressureTransitionPeriod,omitempty"`
 	// Maximum allowed grace period (in seconds) to use when terminating pods in response to a soft eviction threshold being met.
 	EvictionMaxPodGracePeriod int32 `json:"evictionMaxPodGracePeriod,omitempty"`
+	// Maximum number of pods per core. Cannot exceed MaxPods
+	PodsPerCore int32 `json:"podsPerCore"`
 }
 
 type KubeSchedulerConfiguration struct {

pkg/kubelet/kubelet.go

@@ -21,6 +21,7 @@ import (
 	"fmt"
 	"io"
 	"io/ioutil"
+	"math"
 	"net"
 	"net/http"
 	"os"
@@ -216,6 +217,7 @@ func NewMainKubelet(
 	podCIDR string,
 	reconcileCIDR bool,
 	maxPods int,
+	podsPerCore int,
 	nvidiaGPUs int,
 	dockerExecHandler dockertools.ExecHandler,
 	resolverConfig string,
@@ -343,6 +345,7 @@ func NewMainKubelet(
 		nonMasqueradeCIDR:          nonMasqueradeCIDR,
 		reconcileCIDR:              reconcileCIDR,
 		maxPods:                    maxPods,
+		podsPerCore:                podsPerCore,
 		nvidiaGPUs:                 nvidiaGPUs,
 		syncLoopMonitor:            atomic.Value{},
 		resolverConfig:             resolverConfig,
@@ -817,6 +820,9 @@ type Kubelet struct {
 
 	// the list of handlers to call during pod sync.
 	lifecycle.PodSyncHandlers
+
+	// the number of allowed pods per core
+	podsPerCore int
 }
 
 // Validate given node IP belongs to the current host
@@ -3048,8 +3054,13 @@ func (kl *Kubelet) setNodeStatusMachineInfo(node *api.Node) {
 		node.Status.NodeInfo.MachineID = info.MachineID
 		node.Status.NodeInfo.SystemUUID = info.SystemUUID
 		node.Status.Capacity = cadvisor.CapacityFromMachineInfo(info)
-		node.Status.Capacity[api.ResourcePods] = *resource.NewQuantity(
-			int64(kl.maxPods), resource.DecimalSI)
+		if kl.podsPerCore > 0 {
+			node.Status.Capacity[api.ResourcePods] = *resource.NewQuantity(
+				int64(math.Min(float64(info.NumCores*kl.podsPerCore), float64(kl.maxPods))), resource.DecimalSI)
+		} else {
+			node.Status.Capacity[api.ResourcePods] = *resource.NewQuantity(
+				int64(kl.maxPods), resource.DecimalSI)
+		}
 		node.Status.Capacity[api.ResourceNvidiaGPU] = *resource.NewQuantity(
 			int64(kl.nvidiaGPUs), resource.DecimalSI)
 		if node.Status.NodeInfo.BootID != "" &&

pkg/kubemark/hollow_kubelet.go

@@ -43,7 +43,7 @@ func NewHollowKubelet(
 	dockerClient dockertools.DockerInterface,
 	kubeletPort, kubeletReadOnlyPort int,
 	containerManager cm.ContainerManager,
-	maxPods int,
+	maxPods int, podsPerCore int,
 ) *HollowKubelet {
 	testRootDir := integration.MakeTempDirOrDie("hollow-kubelet.", "")
 	manifestFilePath := integration.MakeTempDirOrDie("manifest", testRootDir)
@@ -74,6 +74,7 @@ func NewHollowKubelet(
 			5*time.Minute,           /* OutOfDiskTransitionFrequency */
 			5*time.Minute,           /* EvictionPressureTransitionPeriod */
 			maxPods,
+			podsPerCore,
 			containerManager,
 			nil,
 		),