Community

Blog Events Webinars Tutorials Forum
  1. Blog
  2. Events
  3. Webinars
  4. Tutorials
  5. Forum
Create Account
×
Community Blog Flagger on ASM: Progressive Canary Release Based on Mixerless Telemetry (Part 2) – Application-Level Scaling

Flagger on ASM: Progressive Canary Release Based on Mixerless Telemetry (Part 2) – Application-Level Scaling

Part 2 of this 3-part series describes the configuration of three application-level monitoring metrics in HPA to implement the application-level auto scaling.

Application-level scaling is relative to O&M-level scaling. The utilization ratio of CPU and memory is an application-independent O&M metric, and the HPA configuration for scaling based on such metrics is O&M-level scaling. Metrics, such as request total, request latency, and P99 distribution, are application-related or called business-aware monitoring metrics.

This article describes the configuration of three application-level monitoring metrics in HPA to implement the application-level auto scaling.

Setup HPA

1. Deploy Kube-metrics-adapter

Run the following command to deploy kube-metrics-adapter (For the complete script, please see demo_hpa.sh):

helm --kubeconfig "$USER_CONFIG" -n kube-system install asm-custom-metrics \
  $KUBE_METRICS_ADAPTER_SRC/deploy/charts/kube-metrics-adapter \
  --set prometheus.url=http://prometheus.istio-system.svc:9090

Run the following command to verify the deployment:

#Verify POD
kubectl --kubeconfig "$USER_CONFIG" get po -n kube-system | grep metrics-adapter

asm-custom-metrics-kube-metrics-adapter-6fb4949988-ht8pv   1/1     Running     0          30s

#Verify CRD
kubectl --kubeconfig "$USER_CONFIG" api-versions | grep "autoscaling/v2beta"

autoscaling/v2beta1
autoscaling/v2beta2

#Verify CRD
kubectl --kubeconfig "$USER_CONFIG" get --raw "/apis/external.metrics.k8s.io/v1beta1" | jq .

{
  "kind": "APIResourceList",
  "apiVersion": "v1",
  "groupVersion": "external.metrics.k8s.io/v1beta1",
  "resources": []
}

2. Deploy Loadtester

Run the following command to deploy flagger loadtester:

kubectl --kubeconfig "$USER_CONFIG" apply -f $FLAAGER_SRC/kustomize/tester/deployment.yaml -n test
kubectl --kubeconfig "$USER_CONFIG" apply -f $FLAAGER_SRC/kustomize/tester/service.yaml -n test

3. Deploy HPA

3.1 Scaling Based on the Application Request Total

First, create a configuration of HorizontalPodAutoscaler to sense the application request total (istio_requests_total):

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: podinfo-total
  namespace: test
  annotations:
    metric-config.external.prometheus-query.prometheus/processed-requests-per-second: |
      sum(rate(istio_requests_total{destination_workload_namespace="test",reporter="destination"}[1m]))
spec:
  maxReplicas: 5
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: podinfo
  metrics:
    - type: External
      external:
        metric:
          name: prometheus-query
          selector:
            matchLabels:
              query-name: processed-requests-per-second
        target:
          type: AverageValue
          averageValue: "10"

Run the following command to deploy the HPA configuration:

kubectl --kubeconfig "$USER_CONFIG" apply -f resources_hpa/requests_total_hpa.yaml

Run the following command for verification:

kubectl --kubeconfig "$USER_CONFIG" get --raw "/apis/external.metrics.k8s.io/v1beta1" | jq .

The results are listed below:

{
  "kind": "APIResourceList",
  "apiVersion": "v1",
  "groupVersion": "external.metrics.k8s.io/v1beta1",
  "resources": [
    {
      "name": "prometheus-query",
      "singularName": "",
      "namespaced": true,
      "kind": "ExternalMetricValueList",
      "verbs": [
        "get"
      ]
    }
  ]
}

Similarly, application-level metrics from other dimensions can be used to configure HPA. The examples are listed below:

3.2 Scaling Based on Average Request Latency

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: podinfo-latency-avg
  namespace: test
  annotations:
    metric-config.external.prometheus-query.prometheus/latency-average: |
      sum(rate(istio_request_duration_milliseconds_sum{destination_workload_namespace="test",reporter="destination"}[1m]))
      /sum(rate(istio_request_duration_milliseconds_count{destination_workload_namespace="test",reporter="destination"}[1m]))
spec:
  maxReplicas: 5
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: podinfo
  metrics:
    - type: External
      external:
        metric:
          name: prometheus-query
          selector:
            matchLabels:
              query-name: latency-average
        target:
          type: AverageValue
          averageValue: "0.005"

3.3 Scaling Based on P95 Distribution

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: podinfo-p95
  namespace: test
  annotations:
    metric-config.external.prometheus-query.prometheus/p95-latency: |
      histogram_quantile(0.95,sum(irate(istio_request_duration_milliseconds_bucket{destination_workload_namespace="test",destination_canonical_service="podinfo"}[5m]))by (le))
spec:
  maxReplicas: 5
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: podinfo
  metrics:
    - type: External
      external:
        metric:
          name: prometheus-query
          selector:
            matchLabels:
              query-name: p95-latency
        target:
          type: AverageValue
          averageValue: "4"

Verify the HPA

1. Generate Load

Run the following command to generate experimental traffic to verify if the auto scaling with HPA configuration has taken effect:

alias k="kubectl --kubeconfig $USER_CONFIG"
loadtester=$(k -n test get pod -l "app=flagger-loadtester" -o jsonpath='{.items..metadata.name}')
k -n test exec -it ${loadtester} -c loadtester -- hey -z 5m -c 2 -q 10 http://podinfo:9898

A request lasting for five minutes with a QPS of 10 and a concurrency number of 2 is run on the section below. The detailed hey command is listed below:

Usage: hey [options...] <url>

Options:
  -n  Number of requests to run. Default is 200.
  -c  Number of workers to run concurrently. Total number of requests cannot
      be smaller than the concurrency level. Default is 50.
  -q  Rate limit, in queries per second (QPS) per worker. Default is no rate limit.
  -z  Duration of application to send requests. When duration is reached,
      application stops and exits. If duration is specified, n is ignored.
      Examples: -z 10s -z 3m.
  -o  Output type. If none provided, a summary is printed.
      "csv" is the only supported alternative. Dumps the response
      metrics in comma-separated values format.

  -m  HTTP method, one of GET, POST, PUT, DELETE, HEAD, OPTIONS.
  -H  Custom HTTP header. You can specify as many as needed by repeating the flag.
      For example, -H "Accept: text/html" -H "Content-Type: application/xml" .
  -t  Timeout for each request in seconds. Default is 20, use 0 for infinite.
  -A  HTTP Accept header.
  -d  HTTP request body.
  -D  HTTP request body from file. For example, /home/user/file.txt or ./file.txt.
  -T  Content-type, defaults to "text/html".
  -a  Basic authentication, username:password.
  -x  HTTP Proxy address as host:port.
  -h2 Enable HTTP/2.

  -host HTTP Host header.

  -disable-compression  Disable compression.
  -disable-keepalive    Disable keep-alive, prevents re-use of TCP
                        connections between different HTTP requests.
  -disable-redirects    Disable following of HTTP redirects
  -cpus                 Number of used cpu cores.
                        (default for current machine is 4 cores)

2. Auto Scaling

Run the following command to check the scaling:

watch kubectl --kubeconfig $USER_CONFIG -n test get hpa/podinfo-total

The results are listed below:

Every 2.0s: kubectl --kubeconfig /Users/han/shop_config/ack_zjk -n test get hpa/podinfo                                            East6C16G: Tue Jan 26 18:01:30 2021

NAME      REFERENCE            TARGETS           MINPODS   MAXPODS   REPLICAS   AGE
podinfo   Deployment/podinfo   10056m/10 (avg)   1         5         2          4m45s

It is a similar case with the other two HPAs. The commands are listed below:

kubectl --kubeconfig $USER_CONFIG -n test get hpa

watch kubectl --kubeconfig $USER_CONFIG -n test get hpa/podinfo-latency-avg
watch kubectl --kubeconfig $USER_CONFIG -n test get hpa/podinfo-p95

3. Monitoring Metrics

At the same time, the related real-time application-level metric data can be viewed in Prometheus in real-time, as shown below:

1

Kubernetes cloud native Prometheus Alibaba Cloud Service Mesh Flagger Mixerless Telemetry Progressive Canary Release
0 0 0
Share on

Read previous post:

Flagger on ASM: Progressive Canary Release Based on Mixerless Telemetry (Part 1) – Telemetry Data

Read next post:

Flagger on ASM: Progressive Canary Release Based on Mixerless Telemetry (Part 3) – Progressive Canary Release

feuyeux

6 posts | 0 followers

You may also like

Comments

feuyeux

6 posts | 0 followers

Related Products

More Posts by feuyeux

See All