docs/serving/samples/autoscale-go

Autoscale Sample

A demonstration of the autoscaling capabilities of a Knative Serving Revision.

Prerequisites

1. A Kubernetes cluster with Knative Serving installed.
2. A metrics installation for viewing scaling graphs (optional).
3. Install Docker.
4. Check out the code:

go get -d github.com/knative/docs/serving/samples/autoscale-go

Setup

Build the application container and publish it to a container registry:

1. Move into the sample directory:

   cd $GOPATH/src/github.com/knative/docs
   

2. Set your preferred container registry:

   export REPO="gcr.io/<YOUR_PROJECT_ID>"
   

   • This example shows how to use Google Container Registry (GCR). You will need a Google Cloud Project and to enable the Google Container Registry API.

3. Use Docker to build your application container:

   docker build \
     --tag "${REPO}/serving/samples/autoscale-go" \
     --file=serving/samples/autoscale-go/Dockerfile .
   

4. Push your container to a container registry:

   docker push "${REPO}/serving/samples/autoscale-go"
   

5. Replace the image reference with our published image:

   perl -pi -e \
   "s@github.com/knative/docs/serving/samples/autoscale-go@${REPO}/serving/samples/autoscale-go@g" \
   serving/samples/autoscale-go/service.yaml
   

Deploy the Service

1. Deploy the Knative Serving sample:

   kubectl apply --filename serving/samples/autoscale-go/service.yaml
   

2. Find the ingress hostname and IP and export as an environment variable:

   export IP_ADDRESS=`kubectl get svc knative-ingressgateway --namespace istio-system --output jsonpath="{.status.loadBalancer.ingress[*].ip}"`
   

View the Autoscaling Capabilities

1. Make a request to the autoscale app to see it consume some resources.

   curl --header "Host: autoscale-go.default.example.com" "http://${IP_ADDRESS?}?sleep=100&prime=10000&bloat=5"
   

   Allocated 5 Mb of memory.
   The largest prime less than 10000 is 9973.
   Slept for 100.13 milliseconds.
   

2. Ramp up traffic to maintain 10 in-flight requests.

   go run serving/samples/autoscale-go/test/test.go -sleep 100 -prime 10000 -bloat 5 -qps 9999 -concurrency 300
   

   REQUEST STATS:
   Total: 439      Inflight: 299   Done: 439       Success Rate: 100.00%   Avg Latency: 0.4655 sec
   Total: 1151     Inflight: 245   Done: 712       Success Rate: 100.00%   Avg Latency: 0.4178 sec
   Total: 1706     Inflight: 300   Done: 555       Success Rate: 100.00%   Avg Latency: 0.4794 sec
   Total: 2334     Inflight: 264   Done: 628       Success Rate: 100.00%   Avg Latency: 0.5207 sec
   Total: 2911     Inflight: 300   Done: 577       Success Rate: 100.00%   Avg Latency: 0.4401 sec
   ...
   

   Note: Use CTRL+C to exit the load test.

3. Watch the Knative Serving deployment pod count increase.

   kubectl get deploy --watch
   

   Note: Use CTRL+C to exit watch mode.

Analysis

Algorithm

Knative Serving autoscaling is based on the average number of in-flight requests per pod (concurrency). The system has a default target concurency of 100.0.

For example, if a Revision is receiving 350 requests per second, each of which takes about about .5 seconds, Knative Serving will determine the Revision needs about 2 pods

350 * .5 = 175
175 / 100 = 1.75
ceil(1.75) = 2 pods

Tuning

By default Knative Serving does not limit concurrency in Revision containers. A limit can be set per-Configuration using the ContainerConcurrency field. The autoscaler will target a percentage of ContainerConcurrency instead of the default 100.0.

Dashboards

View the Knative Serving Scaling and Request dashboards (if configured).

kubectl port-forward --namespace monitoring $(kubectl get pods --namespace monitoring --selector=app=grafana --output=jsonpath="{.items..metadata.name}") 3000

Other Experiments

1. Maintain 1000 concurrent requests.

   go run serving/samples/autoscale-go/test/test.go -qps 9999 -concurrency 1000
   

2. Maintain 100 qps with fast requests.

   go run serving/samples/autoscale-go/test/test.go -qps 100 -concurrency 9999
   

3. Maintain 100 qps with slow requests.

   go run serving/samples/autoscale-go/test/test.go -qps 100 -concurrency 9999 -sleep 500
   

4. Heavy CPU usage.

   go run serving/samples/autoscale-go/test/test.go -qps 9999 -concurrency 10 -prime 40000000
   

5. Heavy memory usage.

   go run serving/samples/autoscale-go/test/test.go -qps 9999 -concurrency 5 -bloat 1000
   

Cleanup

kubectl delete --filename serving/samples/autoscale-go/service.yaml

Further reading

1. Autoscaling Developer Documentation