Circuit Breaking

This task shows you how to configure circuit breaking for connections, requests, and outlier detection.

Circuit breaking is an important pattern for creating resilient microservice applications. Circuit breaking allows you to write applications that limit the impact of failures, latency spikes, and other undesirable effects of network peculiarities.

In this task, you will configure circuit breaking rules and then test the configuration by intentionally “tripping” the circuit breaker.

Before you begin

  • Start the httpbin sample.

    If you have enabled automatic sidecar injection, deploy the httpbin service:

    Zip
    $ kubectl apply -f @samples/httpbin/httpbin.yaml@
    

    Otherwise, you have to manually inject the sidecar before deploying the httpbin application:

    Zip
    $ kubectl apply -f <(istioctl kube-inject -f @samples/httpbin/httpbin.yaml@)
    

The httpbin application serves as the backend service for this task.

Configuring the circuit breaker

  1. Create a destination rule to apply circuit breaking settings when calling the httpbin service:

    $ kubectl apply -f - <<EOF
    apiVersion: networking.istio.io/v1alpha3
    kind: DestinationRule
    metadata:
      name: httpbin
    spec:
      host: httpbin
      trafficPolicy:
        connectionPool:
          tcp:
            maxConnections: 1
          http:
            http1MaxPendingRequests: 1
            maxRequestsPerConnection: 1
        outlierDetection:
          consecutiveErrors: 1
          interval: 1s
          baseEjectionTime: 3m
          maxEjectionPercent: 100
    EOF
    
  2. Verify the destination rule was created correctly:

    $ kubectl get destinationrule httpbin -o yaml
    apiVersion: networking.istio.io/v1beta1
    kind: DestinationRule
    ...
    spec:
      host: httpbin
      trafficPolicy:
        connectionPool:
          http:
            http1MaxPendingRequests: 1
            maxRequestsPerConnection: 1
          tcp:
            maxConnections: 1
        outlierDetection:
          baseEjectionTime: 3m
          consecutiveErrors: 1
          interval: 1s
          maxEjectionPercent: 100
    

Adding a client

Create a client to send traffic to the httpbin service. The client is a simple load-testing client called fortio. Fortio lets you control the number of connections, concurrency, and delays for outgoing HTTP calls. You will use this client to “trip” the circuit breaker policies you set in the DestinationRule

  1. Inject the client with the Istio sidecar proxy so network interactions are governed by Istio.

    If you have enabled automatic sidecar injection, deploy the fortio service:

    Zip
    $ kubectl apply -f @samples/httpbin/sample-client/fortio-deploy.yaml@
    

    Otherwise, you have to manually inject the sidecar before deploying the fortio application:

    Zip
    $ kubectl apply -f <(istioctl kube-inject -f @samples/httpbin/sample-client/fortio-deploy.yaml@)
    
  2. Log in to the client pod and use the fortio tool to call httpbin. Pass in -curl to indicate that you just want to make one call:

    $ FORTIO_POD=$(kubectl get pods -lapp=fortio -o 'jsonpath={.items[0].metadata.name}')
    $ kubectl exec -it "$FORTIO_POD" -c fortio -- /usr/bin/fortio load -curl http://httpbin:8000/get
    HTTP/1.1 200 OK
    server: envoy
    date: Tue, 25 Feb 2020 20:25:52 GMT
    content-type: application/json
    content-length: 586
    access-control-allow-origin: *
    access-control-allow-credentials: true
    x-envoy-upstream-service-time: 36
    
    {
      "args": {},
      "headers": {
        "Content-Length": "0",
        "Host": "httpbin:8000",
        "User-Agent": "fortio.org/fortio-1.3.1",
        "X-B3-Parentspanid": "8fc453fb1dec2c22",
        "X-B3-Sampled": "1",
        "X-B3-Spanid": "071d7f06bc94943c",
        "X-B3-Traceid": "86a929a0e76cda378fc453fb1dec2c22",
        "X-Forwarded-Client-Cert": "By=spiffe://cluster.local/ns/default/sa/httpbin;Hash=68bbaedefe01ef4cb99e17358ff63e92d04a4ce831a35ab9a31d3c8e06adb038;Subject=\"\";URI=spiffe://cluster.local/ns/default/sa/default"
      },
      "origin": "127.0.0.1",
      "url": "http://httpbin:8000/get"
    }
    

You can see the request succeeded! Now, it’s time to break something.

Tripping the circuit breaker

In the DestinationRule settings, you specified maxConnections: 1 and http1MaxPendingRequests: 1. These rules indicate that if you exceed more than one connection and request concurrently, you should see some failures when the istio-proxy opens the circuit for further requests and connections.

  1. Call the service with two concurrent connections (-c 2) and send 20 requests (-n 20):

    $ kubectl exec -it "$FORTIO_POD" -c fortio -- /usr/bin/fortio load -c 2 -qps 0 -n 20 -loglevel Warning http://httpbin:8000/get
    20:33:46 I logger.go:97> Log level is now 3 Warning (was 2 Info)
    Fortio 1.3.1 running at 0 queries per second, 6->6 procs, for 20 calls: http://httpbin:8000/get
    Starting at max qps with 2 thread(s) [gomax 6] for exactly 20 calls (10 per thread + 0)
    20:33:46 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:33:47 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:33:47 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    Ended after 59.8524ms : 20 calls. qps=334.16
    Aggregated Function Time : count 20 avg 0.0056869 +/- 0.003869 min 0.000499 max 0.0144329 sum 0.113738
    # range, mid point, percentile, count
    >= 0.000499 <= 0.001 , 0.0007495 , 10.00, 2
    > 0.001 <= 0.002 , 0.0015 , 15.00, 1
    > 0.003 <= 0.004 , 0.0035 , 45.00, 6
    > 0.004 <= 0.005 , 0.0045 , 55.00, 2
    > 0.005 <= 0.006 , 0.0055 , 60.00, 1
    > 0.006 <= 0.007 , 0.0065 , 70.00, 2
    > 0.007 <= 0.008 , 0.0075 , 80.00, 2
    > 0.008 <= 0.009 , 0.0085 , 85.00, 1
    > 0.011 <= 0.012 , 0.0115 , 90.00, 1
    > 0.012 <= 0.014 , 0.013 , 95.00, 1
    > 0.014 <= 0.0144329 , 0.0142165 , 100.00, 1
    # target 50% 0.0045
    # target 75% 0.0075
    # target 90% 0.012
    # target 99% 0.0143463
    # target 99.9% 0.0144242
    Sockets used: 4 (for perfect keepalive, would be 2)
    Code 200 : 17 (85.0 %)
    Code 503 : 3 (15.0 %)
    Response Header Sizes : count 20 avg 195.65 +/- 82.19 min 0 max 231 sum 3913
    Response Body/Total Sizes : count 20 avg 729.9 +/- 205.4 min 241 max 817 sum 14598
    All done 20 calls (plus 0 warmup) 5.687 ms avg, 334.2 qps
    

    It’s interesting to see that almost all requests made it through! The istio-proxy does allow for some leeway.

    Code 200 : 17 (85.0 %)
    Code 503 : 3 (15.0 %)
    
  2. Bring the number of concurrent connections up to 3:

    $ kubectl exec -it "$FORTIO_POD" -c fortio -- /usr/bin/fortio load -c 3 -qps 0 -n 30 -loglevel Warning http://httpbin:8000/get
    20:32:30 I logger.go:97> Log level is now 3 Warning (was 2 Info)
    Fortio 1.3.1 running at 0 queries per second, 6->6 procs, for 30 calls: http://httpbin:8000/get
    Starting at max qps with 3 thread(s) [gomax 6] for exactly 30 calls (10 per thread + 0)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    20:32:30 W http_client.go:679> Parsed non ok code 503 (HTTP/1.1 503)
    Ended after 51.9946ms : 30 calls. qps=576.98
    Aggregated Function Time : count 30 avg 0.0040001633 +/- 0.003447 min 0.0004298 max 0.015943 sum 0.1200049
    # range, mid point, percentile, count
    >= 0.0004298 <= 0.001 , 0.0007149 , 16.67, 5
    > 0.001 <= 0.002 , 0.0015 , 36.67, 6
    > 0.002 <= 0.003 , 0.0025 , 50.00, 4
    > 0.003 <= 0.004 , 0.0035 , 60.00, 3
    > 0.004 <= 0.005 , 0.0045 , 66.67, 2
    > 0.005 <= 0.006 , 0.0055 , 76.67, 3
    > 0.006 <= 0.007 , 0.0065 , 83.33, 2
    > 0.007 <= 0.008 , 0.0075 , 86.67, 1
    > 0.008 <= 0.009 , 0.0085 , 90.00, 1
    > 0.009 <= 0.01 , 0.0095 , 96.67, 2
    > 0.014 <= 0.015943 , 0.0149715 , 100.00, 1
    # target 50% 0.003
    # target 75% 0.00583333
    # target 90% 0.009
    # target 99% 0.0153601
    # target 99.9% 0.0158847
    Sockets used: 20 (for perfect keepalive, would be 3)
    Code 200 : 11 (36.7 %)
    Code 503 : 19 (63.3 %)
    Response Header Sizes : count 30 avg 84.366667 +/- 110.9 min 0 max 231 sum 2531
    Response Body/Total Sizes : count 30 avg 451.86667 +/- 277.1 min 241 max 817 sum 13556
    All done 30 calls (plus 0 warmup) 4.000 ms avg, 577.0 qps
    

    Now you start to see the expected circuit breaking behavior. Only 36.7% of the requests succeeded and the rest were trapped by circuit breaking:

    Code 200 : 11 (36.7 %)
    Code 503 : 19 (63.3 %)
    
  3. Query the istio-proxy stats to see more:

    $ kubectl exec "$FORTIO_POD" -c istio-proxy -- pilot-agent request GET stats | grep httpbin | grep pending
    cluster.outbound|8000||httpbin.default.svc.cluster.local.circuit_breakers.default.rq_pending_open: 0
    cluster.outbound|8000||httpbin.default.svc.cluster.local.circuit_breakers.high.rq_pending_open: 0
    cluster.outbound|8000||httpbin.default.svc.cluster.local.upstream_rq_pending_active: 0
    cluster.outbound|8000||httpbin.default.svc.cluster.local.upstream_rq_pending_failure_eject: 0
    cluster.outbound|8000||httpbin.default.svc.cluster.local.upstream_rq_pending_overflow: 21
    cluster.outbound|8000||httpbin.default.svc.cluster.local.upstream_rq_pending_total: 29
    

    You can see 21 for the upstream_rq_pending_overflow value which means 21 calls so far have been flagged for circuit breaking.

Cleaning up

  1. Remove the rules:

    $ kubectl delete destinationrule httpbin
    
  2. Shutdown the httpbin service and client:

    $ kubectl delete deploy httpbin fortio-deploy
    $ kubectl delete svc httpbin fortio
    
Was this information useful?
Do you have any suggestions for improvement?

Thanks for your feedback!