Kubernetes HPA如何评估内存度量

Question

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
name: backend-hpa
spec:
maxReplicas: 10
minReplicas: 3
scaleTargetRef:
apiVersion: apps/v1
kind: Deployment
name: backend
metrics:
- type: Resource
resource:
name: memory
target:
type: Utilization
averageUtilization: 70
- type: Resource
resource:
name: cpu
target:
type: Utilization
averageUtilization: 70
behavior:
scaleDown:
stabilizationWindowSeconds: 300
policies:
- type: Pods
value: 1
periodSeconds: 100
scaleUp:
stabilizationWindowSeconds: 60
policies:
- type: Pods
value: 2
periodSeconds: 35

我对如何计算内存利用率感到困惑?70%是基于什么阈值?

➜ kd hpa       
Name:                                                     prod-backend-hpa-v1
Namespace:                                                prod
Labels:                                                   argocd.argoproj.io/instance=ccl-backend-prod
Annotations:                                              <none>
CreationTimestamp:                                        Mon, 13 Sep 2021 17:39:44 -0700
Reference:                                                Deployment/prod-backend-v1
Metrics:                                                  ( current / target )
resource memory on pods  (as a percentage of request):  31% (85408426666m) / 70%
resource cpu on pods  (as a percentage of request):     0% (1m) / 70%
Min replicas:                                             3
Max replicas:                                             10
Behavior:
Scale Up:
Stabilization Window: 60 seconds
Select Policy: Max
Policies:
- Type: Pods  Value: 2  Period: 35 seconds
Scale Down:
Stabilization Window: 300 seconds
Select Policy: Max
Policies:
- Type: Pods  Value: 1  Period: 100 seconds
Deployment pods:    3 current / 3 desired
Conditions:
Type            Status  Reason            Message
----            ------  ------            -------
AbleToScale     True    ReadyForNewScale  recommended size matches current size
ScalingActive   True    ValidMetricFound  the HPA was able to successfully calculate a replica count from memory resource utilization (percentage of request)
ScalingLimited  True    TooFewReplicas    the desired replica count is less than the minimum replica count
Events:           <none>

resource memory on pods  (as a percentage of request):  31% (85408426666m) / 70%

计算?

我的部署有以下请求/限制设置

Containers:
backend:
Port:       8080/TCP
Host Port:  0/TCP
Command:
web
Limits:
cpu:     1
memory:  1Gi
Requests:
cpu:      200m
memory:   256Mi

my current pods

➜ k top po
W0914 13:38:13.793286 3006942 top_pod.go:140] Using json format to get metrics. Next release will switch to protocol-buffers, switch early by passing --use-protocol-buffers flag
NAME                              CPU(cores)   MEMORY(bytes)   
prod-backend-v1-7858bddc4-5t8r5   1m           80Mi            
prod-backend-v1-7858bddc4-67llh   1m           81Mi            
prod-backend-v1-7858bddc4-98wj2   1m           82Mi 

所以似乎31%被计算为81/256 ~31%.

但是这样做正确吗?

我认为期望81/LIMIT = 81/1024 = ~8%.

在逻辑上是正确的

Answer 1

type=Utilization &&averageUtilization:70

是所有相关pod中资源度量的平均值的目标值，表示为请求值的百分比。为pod提供资源。当前仅对资源度量源类型

有效

type = AverageValue,,averageValue: 500Mi

averageValue是所有相关pod(作为数量)的度量平均值的目标值

所以我的HPA内存指标变成了:

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
name: backend-hpa
spec:
maxReplicas: 10
minReplicas: 3
scaleTargetRef:
apiVersion: apps/v1
kind: Deployment
name: backend
metrics:
- type: Resource
resource:
name: memory
target:
type: AverageValue
averageValue: 500Mi # averageValue is the target value of the average of the metric across all relevant pods (as a quantity), https://www.pulumi.com/docs/reference/pkg/kubernetes/autoscaling/v2beta2/horizontalpodautoscalerlist/#metrictarget
- type: Resource
resource:
name: cpu
target:
type: Utilization
averageUtilization: 70 # represented as a percentage of the requested value of the resource for the pods.
behavior:
scaleDown:
stabilizationWindowSeconds: 300
policies:
- type: Pods
value: 1
periodSeconds: 100
scaleUp:
stabilizationWindowSeconds: 60
policies:
- type: Pods
value: 2
periodSeconds: 35

更多信息:https://www.pulumi.com/docs/reference/pkg/kubernetes/autoscaling/v2beta2/horizontalpodautoscalerlist/metrictarget