MeshFaultInjection

Overview

With the MeshFaultInjection policy you can easily test your microservices against resiliency. You can:

• return specific HTTP error codes to verify error-handling logic.
• introduce artificial response delays to test timeout and retry behavior.
• restrict response speed to emulate slow or unstable network connections.

With matches api you can select subset of clients for which faults should be injected.

Configuration

MeshFaultInjection supports three types of faults

Abort

Abort defines a configuration of not delivering requests to destination service and replacing the responses from destination data plane by predefined status code.

• httpStatus - HTTP status code which will be returned to source side, has to be in [100 - 599] range
• percentage - a percentage of requests on which abort will be injected, has to be in [0.0 - 100.0] range. If the value is a double number, put it in quotes.

Delay

Delay defines a configuration of delaying a response from a destination.

• value - the duration during which the response will be delayed
• percentage - a percentage of requests on which abort will be injected, has to be in [0.0 - 100.0] range. If the value is a double number, put it in quotes.

ResponseBandwidth limit

ResponseBandwidth defines a configuration to limit the speed of responding to requests.

• limit - represented by value measure in Gbps, Mbps, kbps, or bps, for example 10kbps
• percentage - a percentage of requests on which abort will be injected, has to be in [0.0 - 100.0] range. If the value is a double number, put it in quotes.

List of multiple failures

MeshFaultInjection allows configuring a list of HTTP faults. They execute in the same order as they were defined.

default:
  http:
    - abort:
        httpStatus: 500
        percentage: "2.5"
      delay:
        value: 5s
        percentage: 5
      responseBandwidth:
        limit: "50Mbps"
        percentage: 50
    - abort:
        httpStatus: 500
        percentage: 10
    - delay:
        value: 5s
        percentage: 5

It’s worth mentioning that percentage of the next filter depends on the percentage of previous ones.

http:
  - abort:
      httpStatus: 500
      percentage: 70
  - abort:
      httpStatus: 503
      percentage: 50

That means that for 70% of requests, it returns 500 and for 50% of the 30% that passed it returns 503.

Examples

Service backend returns 500 for 50% of requests from frontend service

apiVersion: kuma.io/v1alpha1
kind: MeshFaultInjection
metadata:
  name: default-fault-injection
  namespace: kuma-demo
  labels:
    kuma.io/mesh: default
spec:
  targetRef:
    kind: Dataplane
    labels:
      app: backend
  rules:
  - matches:
    - spiffeID:
        type: Exact
        value: spiffe://my-mesh.us-east-2.mesh.local/ns/test/sa/frontend
    default:
      http:
      - abort:
          httpStatus: 500
          percentage: 50

type: MeshFaultInjection
mesh: default
name: default-fault-injection
spec:
  targetRef:
    kind: Dataplane
    labels:
      app: backend
  rules:
  - matches:
    - spiffeID:
        type: Exact
        value: spiffe://my-mesh.us-east-2.mesh.local/ns/test/sa/frontend
    default:
      http:
      - abort:
          httpStatus: 500
          percentage: 50

50.5% of requests to service backend from any frontend service is going to be delayed by 5 seconds

apiVersion: kuma.io/v1alpha1
kind: MeshFaultInjection
metadata:
  name: default-fault-injection
  namespace: kuma-demo
  labels:
    kuma.io/mesh: default
spec:
  targetRef:
    kind: Dataplane
    labels:
      app: backend
  rules:
  - matches:
    - spiffeID:
        type: Exact
        value: spiffe://my-mesh.us-east-2.mesh.local/ns/test/sa/frontend
    default:
      http:
      - delay:
          percentage: '50.5'
          value: 5s

type: MeshFaultInjection
mesh: default
name: default-fault-injection
spec:
  targetRef:
    kind: Dataplane
    labels:
      app: backend
  rules:
  - matches:
    - spiffeID:
        type: Exact
        value: spiffe://my-mesh.us-east-2.mesh.local/ns/test/sa/frontend
    default:
      http:
      - delay:
          percentage: '50.5'
          value: 5s

Backend service applies all faults from list for requests coming from frontend service

apiVersion: kuma.io/v1alpha1
kind: MeshFaultInjection
metadata:
  name: default-fault-injection
  namespace: kuma-demo
  labels:
    kuma.io/mesh: default
spec:
  targetRef:
    kind: Dataplane
    labels:
      app: backend
  rules:
  - matches:
    - spiffeID:
        type: Exact
        value: spiffe://my-mesh.us-east-2.mesh.local/ns/test/sa/frontend
    default:
      http:
      - abort:
          httpStatus: 500
          percentage: '2.5'
      - abort:
          httpStatus: 503
          percentage: 10
      - delay:
          value: 5s
          percentage: 5

type: MeshFaultInjection
mesh: default
name: default-fault-injection
spec:
  targetRef:
    kind: Dataplane
    labels:
      app: backend
  rules:
  - matches:
    - spiffeID:
        type: Exact
        value: spiffe://my-mesh.us-east-2.mesh.local/ns/test/sa/frontend
    default:
      http:
      - abort:
          httpStatus: 500
          percentage: '2.5'
      - abort:
          httpStatus: 503
          percentage: 10
      - delay:
          value: 5s
          percentage: 5

All policy options