koordlet: add net qos plugin

Signed-off-by: l1b0k <libokang.dev@gmail.com>
koordinator-sh · Feb 18, 2024 · d8d35d6 · d8d35d6
1 parent 7c7a844
commit d8d35d6
Show file tree

Hide file tree

Showing 10 changed files with 630 additions and 0 deletions.
diff --git a/apis/extension/constants.go b/apis/extension/constants.go
@@ -40,6 +40,10 @@ const (
 	// LabelPodMutatingUpdate is a label key that pods with `pod.koordinator.sh/mutating-update=true` will
 	// be mutated by Koordinator webhook when updating.
 	LabelPodMutatingUpdate = PodDomainPrefix + "/mutating-update"
+
+	// AnnotationNetworkQOS are used to set bandwidth for Pod. The unit is bps.
+	// For example, 10M means 10 megabits per second.
+	AnnotationNetworkQOS = DomainPrefix + "networkQOS"
 )
 
 type AggregationType string

diff --git a/apis/slo/v1alpha1/nodeslo_types.go b/apis/slo/v1alpha1/nodeslo_types.go
@@ -243,6 +243,8 @@ type NetworkQOS struct {
 type ResourceQOSPolicies struct {
 	// applied policy for the CPU QoS, default = "groupIdentity"
 	CPUPolicy *CPUQOSPolicy `json:"cpuPolicy,omitempty"`
+
+	NetworkQOS *NetworkQOSCfg `json:"networkQOS,omitempty"`
 }
 
 type ResourceQOSStrategy struct {

diff --git a/apis/slo/v1alpha1/zz_generated.deepcopy.go b/apis/slo/v1alpha1/zz_generated.deepcopy.go
diff --git a/config/crd/bases/slo.koordinator.sh_nodeslos.yaml b/config/crd/bases/slo.koordinator.sh_nodeslos.yaml
@@ -1222,6 +1222,59 @@ spec:
                       cpuPolicy:
                         description: applied policy for the CPU QoS, default = "groupIdentity"
                         type: string
+                      networkQOS:
+                        properties:
+                          egressLimit:
+                            anyOf:
+                            - type: integer
+                            - type: string
+                            default: 100
+                            description: 'EgressLimit describes the maximum network
+                              bandwidth can be used in the egress direction, unit:
+                              bps(bytes per second), two expressions are supported，int
+                              and string, int: percentage based on total bandwidth，valid
+                              in 0-100 string: a specific network bandwidth value,
+                              eg: 50M.'
+                            x-kubernetes-int-or-string: true
+                          egressRequest:
+                            anyOf:
+                            - type: integer
+                            - type: string
+                            default: 0
+                            description: 'EgressRequest describes the minimum network
+                              bandwidth guaranteed in the egress direction. unit:
+                              bps(bytes per second), two expressions are supported，int
+                              and string, int: percentage based on total bandwidth，valid
+                              in 0-100 string: a specific network bandwidth value,
+                              eg: 50M.'
+                            x-kubernetes-int-or-string: true
+                          enable:
+                            type: boolean
+                          ingressLimit:
+                            anyOf:
+                            - type: integer
+                            - type: string
+                            default: 100
+                            description: 'IngressLimit describes the maximum network
+                              bandwidth can be used in the ingress direction, unit:
+                              bps(bytes per second), two expressions are supported，int
+                              and string, int: percentage based on total bandwidth，valid
+                              in 0-100 string: a specific network bandwidth value,
+                              eg: 50M.'
+                            x-kubernetes-int-or-string: true
+                          ingressRequest:
+                            anyOf:
+                            - type: integer
+                            - type: string
+                            default: 0
+                            description: 'IngressRequest describes the minimum network
+                              bandwidth guaranteed in the ingress direction. unit:
+                              bps(bytes per second), two expressions are supported，int
+                              and string, int: percentage based on total bandwidth，valid
+                              in 0-100 string: a specific network bandwidth value,
+                              eg: 50M.'
+                            x-kubernetes-int-or-string: true
+                        type: object
                     type: object
                   systemClass:
                     description: ResourceQOS for system pods

diff --git a/pkg/koordlet/runtimehooks/config.go b/pkg/koordlet/runtimehooks/config.go
@@ -31,6 +31,7 @@ import (
 	"github.com/koordinator-sh/koordinator/pkg/koordlet/runtimehooks/hooks/cpuset"
 	"github.com/koordinator-sh/koordinator/pkg/koordlet/runtimehooks/hooks/gpu"
 	"github.com/koordinator-sh/koordinator/pkg/koordlet/runtimehooks/hooks/groupidentity"
+	"github.com/koordinator-sh/koordinator/pkg/koordlet/runtimehooks/hooks/terwayqos"
 	"github.com/koordinator-sh/koordinator/pkg/koordlet/util/system"
 )
 
@@ -73,6 +74,11 @@ const (
 	// owner: @saintube @zwzhang0107
 	// alpha: v1.4
 	CoreSched featuregate.Feature = "CoreSched"
+
+	// TerwayQoS enables net QoS feature of koordlet.
+	// owner: @l1b0k
+	// alpha: v1.5
+	TerwayQoS featuregate.Feature = "TerwayQoS"
 )
 
 var (
@@ -83,6 +89,7 @@ var (
 		BatchResource:    {Default: true, PreRelease: featuregate.Beta},
 		CPUNormalization: {Default: false, PreRelease: featuregate.Alpha},
 		CoreSched:        {Default: false, PreRelease: featuregate.Alpha},
+		TerwayQoS:        {Default: false, PreRelease: featuregate.Alpha},
 	}
 
 	runtimeHookPlugins = map[featuregate.Feature]HookPlugin{
@@ -92,6 +99,7 @@ var (
 		BatchResource:    batchresource.Object(),
 		CPUNormalization: cpunormalization.Object(),
 		CoreSched:        coresched.Object(),
+		TerwayQoS:        terwayqos.Object(),
 	}
 )