Injecting Faults

It is easy to inject failures into applications by using the Traffic SplitAPI of theService Mesh Interface. TrafficSplit allows you toredirect a percentage of traffic to a specific backend. This backend iscompletely flexible and can return whatever responses you want - 500s, timeoutsor even crazy payloads.

The books demo is a great way to show off this behavior. Theoverall topology looks like:

)

In this guide, you will split some of the requests from webapp to books.Most requests will end up at the correct books destination, however some ofthem will be redirected to a faulty backend. This backend will return 500s forevery request and inject faults into the webapp service. No code changes arerequired and as this method is configuration driven, it is a process that can beadded to integration tests and CI pipelines. If you are really living the chaosengineering lifestyle, fault injection could even be used in production.

Prerequisites

To use this guide, you’ll need to have Linkerd installed on your cluster.Follow the Installing Linkerd Guide if you haven’t alreadydone this.

Setup the service

First, add the books sample application to your cluster:

kubectl create ns booksapp && \
  linkerd inject https://run.linkerd.io/booksapp.yml | \
  kubectl -n booksapp apply -f -

As this manifest is used as a demo elsewhere, it has been configured with anerror rate. To show how fault injection works, the error rate needs to beremoved so that there is a reliable baseline. To increase success rate forbooksapp to 100%, run:

kubectl -n booksapp patch deploy authors \
  --type='json' \
  -p='[{"op":"remove", "path":"/spec/template/spec/containers/0/env/2"}]'

After a little while, the stats will show 100% success rate. You can verify thisby running:

linkerd -n booksapp stat deploy

The output will end up looking at little like:

NAME      MESHED   SUCCESS      RPS   LATENCY_P50   LATENCY_P95   LATENCY_P99   TCP_CONN
authors      1/1   100.00%   7.1rps           4ms          26ms          33ms          6
books        1/1   100.00%   8.6rps           6ms          73ms          95ms          6
traffic      1/1         -        -             -             -             -          -
webapp       3/3   100.00%   7.9rps          20ms          76ms          95ms          9

Create the faulty backend

Injecting faults into booksapp requires a service that is configured to returnerrors. To do this, you can start NGINX and configure it to return 500s byrunning:

cat <<EOF | linkerd inject - | kubectl apply -f -
apiVersion: v1
kind: ConfigMap
metadata:
  name: error-injector
  namespace: booksapp
data:
 nginx.conf: |-
    events {}
    http {
        server {
          listen 8080;
            location / {
                return 500;
            }
        }
    }
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: error-injector
  namespace: booksapp
  labels:
    app: error-injector
spec:
  selector:
    matchLabels:
      app: error-injector
  replicas: 1
  template:
    metadata:
      labels:
        app: error-injector
    spec:
      containers:
        - name: nginx
          image: nginx:alpine
          volumeMounts:
            - name: nginx-config
              mountPath: /etc/nginx/nginx.conf
              subPath: nginx.conf
      volumes:
        - name: nginx-config
          configMap:
            name: error-injector
---
apiVersion: v1
kind: Service
metadata:
  name: error-injector
  namespace: booksapp
spec:
  ports:
  - name: service
    port: 8080
  selector:
    app: error-injector
EOF

Inject faults

With booksapp and NGINX running, it is now time to partially split the trafficbetween an existing backend, books, and the newly createderror-injector. This is done by adding aTrafficSplitconfiguration to your cluster:

cat <<EOF | kubectl apply -f -
apiVersion: split.smi-spec.io/v1alpha1
kind: TrafficSplit
metadata:
  name: error-split
  namespace: booksapp
spec:
  service: books
  backends:
  - service: books
    weight: 900m
  - service: error-injector
    weight: 100m
EOF

When Linkerd sees traffic going to the books service, it will send ⁹⁄₁₀requests to the original service and ¹⁄₁₀ to the error injector. You can seewhat this looks like by running stat and filtering explicitly to just therequests from webapp:

linkerd -n booksapp routes deploy/webapp --to service/books

Unlike the previous stat command which only looks at the requests received byservers, this routes command filters to all the requests being issued bywebapp destined for the books service itself. The output should show a 90%success rate:

ROUTE       SERVICE   SUCCESS      RPS   LATENCY_P50   LATENCY_P95   LATENCY_P99
[DEFAULT]     books    90.08%   2.0rps           5ms          69ms          94ms

NoteIn this instance, you are looking at the service instead of the deployment. Ifyou were to run this command and look at deploy/books, the success rate wouldstill be 100%. The reason for this is that error-injector is a completelyseparate deployment and traffic is being shifted at the service level. Therequests never reach the books pods and are instead rerouted to the errorinjector’s pods.

Cleanup

To remove everything in this guide from your cluster, run:

kubectl delete ns booksapp