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

Move GPU metrics for request, limit, usage & reserved_capacity into pod store #225

Merged
merged 3 commits into from
Aug 14, 2024
Merged

Move GPU metrics for request, limit, usage & reserved_capacity into pod store #225

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9c631d4
Move GPU request, limit, usage and reserved_capacity metrics into pod…
sky333999 Aug 10, 2024
bab8231
Update test
sky333999 Aug 14, 2024
959860c
Minor tweaks
sky333999 Aug 14, 2024
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
14 changes: 8 additions & 6 deletions internal/aws/containerinsight/const.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -141,9 +141,10 @@ const (
EfaRxDropped = "rx_dropped"
EfaTxBytes = "tx_bytes"

GpuLimit = "gpu_limit"
GpuTotal = "gpu_total"
GpuRequest = "gpu_request"
GpuLimit = "gpu_limit"
GpuUsageTotal = "gpu_usage_total"
GpuRequest = "gpu_request"
GpuReservedCapacity = "gpu_reserved_capacity"

// Define the metric types
TypeCluster = "Cluster"
Expand Down Expand Up @@ -325,8 +326,9 @@ func init() {
EfaRxDropped: UnitCountPerSec,
EfaTxBytes: UnitBytesPerSec,

GpuLimit: UnitCount,
GpuTotal: UnitCount,
GpuRequest: UnitCount,
GpuLimit: UnitCount,
GpuUsageTotal: UnitCount,
GpuRequest: UnitCount,
GpuReservedCapacity: UnitPercent,
}
}
50 changes: 16 additions & 34 deletions internal/aws/containerinsight/utils_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -431,6 +431,10 @@ func TestConvertToOTLPMetricsForNodeMetrics(t *testing.T) {
"node_network_tx_packets": 37.802748111760124,
"node_number_of_running_containers": int32(7),
"node_number_of_running_pods": int64(7),
"node_gpu_request": int32(7),
"node_gpu_limit": int32(7),
"node_gpu_usage_total": int32(7),
"node_gpu_reserved_capacity": 3.0093851356081194,
}
expectedUnits = map[string]string{
"node_cpu_limit": "",
Expand Down Expand Up @@ -467,6 +471,10 @@ func TestConvertToOTLPMetricsForNodeMetrics(t *testing.T) {
"node_network_tx_packets": UnitCountPerSec,
"node_number_of_running_containers": UnitCount,
"node_number_of_running_pods": UnitCount,
"node_gpu_request": UnitCount,
"node_gpu_limit": UnitCount,
"node_gpu_usage_total": UnitCount,
"node_gpu_reserved_capacity": UnitPercent,
}
tags = map[string]string{
"AutoScalingGroupName": "eks-a6bb9db9-267c-401c-db55-df8ef645b06f",
Expand Down Expand Up @@ -716,6 +724,10 @@ func TestConvertToOTLPMetricsForPodMetrics(t *testing.T) {
"pod_status_unknown": 0,
"pod_status_ready": 1,
"pod_status_scheduled": 1,
"pod_gpu_request": 1,
"pod_gpu_limit": 1,
"pod_gpu_usage_total": 1,
"pod_gpu_reserved_capacity": 2.3677681271483983,
}
expectedUnits = map[string]string{
"pod_cpu_limit": "",
Expand Down Expand Up @@ -762,6 +774,10 @@ func TestConvertToOTLPMetricsForPodMetrics(t *testing.T) {
"pod_status_unknown": UnitCount,
"pod_status_ready": UnitCount,
"pod_status_scheduled": UnitCount,
"pod_gpu_request": UnitCount,
"pod_gpu_limit": UnitCount,
"pod_gpu_usage_total": UnitCount,
"pod_gpu_reserved_capacity": UnitPercent,
}
tags = map[string]string{
"ClusterName": "eks-aoc",
Expand Down Expand Up @@ -913,37 +929,3 @@ func TestConvertToOTLPMetricsForPodEfaMetrics(t *testing.T) {
md = ConvertToOTLPMetrics(fields, tags, zap.NewNop())
checkMetricsAreExpected(t, md, fields, tags, expectedUnits)
}

func TestConvertToOTLPMetricsForAcceleratorCountMetrics(t *testing.T) {
var fields map[string]any
var expectedUnits map[string]string
var tags map[string]string
var md pmetric.Metrics
now := time.Now()
timestamp := strconv.FormatInt(now.UnixNano(), 10)

fields = map[string]any{
"pod_gpu_limit": int64(3),
"pod_gpu_total": int64(3),
"pod_gpu_request": int64(3),
}
expectedUnits = map[string]string{
"pod_gpu_limit": UnitCount,
"pod_gpu_total": UnitCount,
"pod_gpu_request": UnitCount,
}
tags = map[string]string{
"ClusterName": "eks-aoc",
"InstanceId": "i-01bf9fb097cbf3205",
"InstanceType": "t2.xlarge",
"Namespace": "amazon-cloudwatch",
"NodeName": "ip-192-168-12-170.ec2.internal",
"PodName": "cloudwatch-agent",
"ContainerName": "cloudwatch-agent",
"Type": "PodGPU",
"Version": "0",
"Timestamp": timestamp,
}
md = ConvertToOTLPMetrics(fields, tags, zap.NewNop())
checkMetricsAreExpected(t, md, fields, tags, expectedUnits)
}
53 changes: 12 additions & 41 deletions internal/aws/k8s/k8sclient/node.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,11 +18,6 @@ import (
"k8s.io/client-go/tools/cache"
)

const (
instanceTypeLabelKey = "node.kubernetes.io/instance-type"
instanceTypeLabelKeyBeta = "beta.kubernetes.io/instance-type"
)

// This needs to be reviewed for newer versions of k8s.
var failedNodeConditions = map[v1.NodeConditionType]bool{
v1.NodeMemoryPressure: true,
Expand All @@ -32,7 +27,6 @@ var failedNodeConditions = map[v1.NodeConditionType]bool{
}

type NodeClient interface {
NodeInfos() map[string]*NodeInfo
// Get the number of failed nodes for current cluster
ClusterFailedNodeCount() int
// Get the number of nodes for current cluster
Expand Down Expand Up @@ -78,23 +72,13 @@ type nodeClient struct {
captureNodeLevelInfo bool

mu sync.RWMutex
nodeInfos map[string]*NodeInfo
clusterFailedNodeCount int
clusterNodeCount int
nodeToCapacityMap map[string]v1.ResourceList
nodeToAllocatableMap map[string]v1.ResourceList
nodeToConditionsMap map[string]map[v1.NodeConditionType]v1.ConditionStatus
}

func (c *nodeClient) NodeInfos() map[string]*NodeInfo {
if c.store.GetResetRefreshStatus() {
c.refresh()
}
c.mu.RLock()
defer c.mu.RUnlock()
return c.nodeInfos
}

func (c *nodeClient) ClusterFailedNodeCount() int {
if c.store.GetResetRefreshStatus() {
c.refresh()
Expand Down Expand Up @@ -161,26 +145,24 @@ func (c *nodeClient) refresh() {
nodeToAllocatableMap := make(map[string]v1.ResourceList)
nodeToConditionsMap := make(map[string]map[v1.NodeConditionType]v1.ConditionStatus)

nodeInfos := map[string]*NodeInfo{}
for _, obj := range objsList {
node := obj.(*NodeInfo)
nodeInfos[node.Name] = node
node := obj.(*nodeInfo)

if c.captureNodeLevelInfo {
nodeToCapacityMap[node.Name] = node.Capacity
nodeToAllocatableMap[node.Name] = node.Allocatable
nodeToCapacityMap[node.name] = node.capacity
nodeToAllocatableMap[node.name] = node.allocatable
conditionsMap := make(map[v1.NodeConditionType]v1.ConditionStatus)
for _, condition := range node.Conditions {
for _, condition := range node.conditions {
conditionsMap[condition.Type] = condition.Status
}
nodeToConditionsMap[node.Name] = conditionsMap
nodeToConditionsMap[node.name] = conditionsMap
}
clusterNodeCountNew++

failed := false

Loop:
for _, condition := range node.Conditions {
for _, condition := range node.conditions {
if _, ok := failedNodeConditions[condition.Type]; ok {
// match the failedNodeConditions type we care about
if condition.Status != v1.ConditionFalse {
Expand All @@ -196,7 +178,6 @@ func (c *nodeClient) refresh() {
}
}

c.nodeInfos = nodeInfos
c.clusterFailedNodeCount = clusterFailedNodeCountNew
c.clusterNodeCount = clusterNodeCountNew
c.nodeToCapacityMap = nodeToCapacityMap
Expand Down Expand Up @@ -241,23 +222,13 @@ func transformFuncNode(obj any) (any, error) {
if !ok {
return nil, fmt.Errorf("input obj %v is not Node type", obj)
}
info := new(NodeInfo)
info.Name = node.Name
info.Capacity = node.Status.Capacity
info.Allocatable = node.Status.Allocatable
info.Conditions = []*NodeCondition{}
info.ProviderID = node.Spec.ProviderID
if instanceType, ok := node.Labels[instanceTypeLabelKey]; ok {
info.InstanceType = instanceType
} else {
// fallback for compatibility with k8s versions older than v1.17
// https://kubernetes.io/docs/reference/labels-annotations-taints/#beta-kubernetes-io-instance-type-deprecated
if instanceType, ok := node.Labels[instanceTypeLabelKeyBeta]; ok {
info.InstanceType = instanceType
}
}
info := new(nodeInfo)
info.name = node.Name
info.capacity = node.Status.Capacity
info.allocatable = node.Status.Allocatable
info.conditions = []*NodeCondition{}
for _, condition := range node.Status.Conditions {
info.Conditions = append(info.Conditions, &NodeCondition{
info.conditions = append(info.conditions, &NodeCondition{
Type: condition.Type,
Status: condition.Status,
})
Expand Down
12 changes: 5 additions & 7 deletions internal/aws/k8s/k8sclient/node_info.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,13 +7,11 @@ import (
v1 "k8s.io/api/core/v1"
)

type NodeInfo struct {
Name string
Conditions []*NodeCondition
Capacity v1.ResourceList
Allocatable v1.ResourceList
ProviderID string
InstanceType string
type nodeInfo struct {
name string
conditions []*NodeCondition
capacity v1.ResourceList
allocatable v1.ResourceList
}

type NodeCondition struct {
Expand Down
Loading
Loading