NGINX Canary Deployments

This guide shows you how to use the NGINX ingress controller and Flagger to automate canary deployments and A/B testing.

Flagger NGINX Ingress Controller

Prerequisites

Flagger requires a Kubernetes cluster v1.11 or newer and NGINX ingress 0.24 or newer.

Install NGINX with Helm:

helm upgrade -i nginx-ingress stable/nginx-ingress \
--namespace ingress-nginx \
--set controller.stats.enabled=true \
--set controller.metrics.enabled=true \
--set controller.podAnnotations."prometheus\.io/scrape"=true \
--set controller.podAnnotations."prometheus\.io/port"=10254

Install Flagger and the Prometheus add-on in the same namespace as NGINX:

helm repo add flagger https://flagger.app
helm upgrade -i flagger flagger/flagger \
--namespace ingress-nginx \
--set prometheus.install=true \
--set meshProvider=nginx

Optionally you can enable Slack notifications:

helm upgrade -i flagger flagger/flagger \
--reuse-values \
--namespace ingress-nginx \
--set slack.url=https://hooks.slack.com/services/YOUR/SLACK/WEBHOOK \
--set slack.channel=general \
--set slack.user=flagger

Bootstrap

Flagger takes a Kubernetes deployment and optionally a horizontal pod autoscaler (HPA), then creates a series of objects (Kubernetes deployments, ClusterIP services and canary ingress). These objects expose the application outside the cluster and drive the canary analysis and promotion.

Create a test namespace:

kubectl create ns test

Create a deployment and a horizontal pod autoscaler:

kubectl apply -f ${REPO}/artifacts/nginx/deployment.yaml
kubectl apply -f ${REPO}/artifacts/nginx/hpa.yaml

Deploy the load testing service to generate traffic during the canary analysis:

helm upgrade -i flagger-loadtester flagger/loadtester \
--namespace=test

Create an ingress definition (replace app.example.com with your own domain):

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
name: podinfo
namespace: test
labels:
app: podinfo
annotations:
kubernetes.io/ingress.class: "nginx"
spec:
rules:
- host: app.example.com
http:
paths:
- backend:
serviceName: podinfo
servicePort: 9898

Save the above resource as podinfo-ingress.yaml and then apply it:

kubectl apply -f ./podinfo-ingress.yaml

Create a canary custom resource (replace app.example.com with your own domain):

apiVersion: flagger.app/v1alpha3
kind: Canary
metadata:
name: podinfo
namespace: test
spec:
provider: nginx
# deployment reference
targetRef:
apiVersion: apps/v1
kind: Deployment
name: podinfo
# ingress reference
ingressRef:
apiVersion: extensions/v1beta1
kind: Ingress
name: podinfo
# HPA reference (optional)
autoscalerRef:
apiVersion: autoscaling/v2beta1
kind: HorizontalPodAutoscaler
name: podinfo
# the maximum time in seconds for the canary deployment
# to make progress before it is rollback (default 600s)
progressDeadlineSeconds: 60
service:
# container port
port: 9898
canaryAnalysis:
# schedule interval (default 60s)
interval: 10s
# max number of failed metric checks before rollback
threshold: 10
# max traffic percentage routed to canary
# percentage (0-100)
maxWeight: 50
# canary increment step
# percentage (0-100)
stepWeight: 5
# NGINX Prometheus checks
metrics:
- name: request-success-rate
# minimum req success rate (non 5xx responses)
# percentage (0-100)
threshold: 99
interval: 1m
# load testing (optional)
webhooks:
- name: load-test
url: http://flagger-loadtester.test/
timeout: 5s
metadata:
type: cmd
cmd: "hey -z 1m -q 10 -c 2 http://app.example.com/"

Save the above resource as podinfo-canary.yaml and then apply it:

kubectl apply -f ./podinfo-canary.yaml

After a couple of seconds Flagger will create the canary objects:

# applied
deployment.apps/podinfo
horizontalpodautoscaler.autoscaling/podinfo
ingresses.extensions/podinfo
canary.flagger.app/podinfo
# generated
deployment.apps/podinfo-primary
horizontalpodautoscaler.autoscaling/podinfo-primary
service/podinfo
service/podinfo-canary
service/podinfo-primary
ingresses.extensions/podinfo-canary

Automated canary promotion

Flagger implements a control loop that gradually shifts traffic to the canary while measuring key performance indicators like HTTP requests success rate, requests average duration and pod health. Based on analysis of the KPIs a canary is promoted or aborted, and the analysis result is published to Slack.

Flagger Canary Stages

Trigger a canary deployment by updating the container image:

kubectl -n test set image deployment/podinfo \
podinfod=stefanprodan/podinfo:2.0.1

Flagger detects that the deployment revision changed and starts a new rollout:

kubectl -n test describe canary/podinfo
Status:
Canary Weight: 0
Failed Checks: 0
Phase: Succeeded
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Synced 3m flagger New revision detected podinfo.test
Normal Synced 3m flagger Scaling up podinfo.test
Warning Synced 3m flagger Waiting for podinfo.test rollout to finish: 0 of 1 updated replicas are available
Normal Synced 3m flagger Advance podinfo.test canary weight 5
Normal Synced 3m flagger Advance podinfo.test canary weight 10
Normal Synced 3m flagger Advance podinfo.test canary weight 15
Normal Synced 2m flagger Advance podinfo.test canary weight 20
Normal Synced 2m flagger Advance podinfo.test canary weight 25
Normal Synced 1m flagger Advance podinfo.test canary weight 30
Normal Synced 1m flagger Advance podinfo.test canary weight 35
Normal Synced 55s flagger Advance podinfo.test canary weight 40
Normal Synced 45s flagger Advance podinfo.test canary weight 45
Normal Synced 35s flagger Advance podinfo.test canary weight 50
Normal Synced 25s flagger Copying podinfo.test template spec to podinfo-primary.test
Warning Synced 15s flagger Waiting for podinfo-primary.test rollout to finish: 1 of 2 updated replicas are available
Normal Synced 5s flagger Promotion completed! Scaling down podinfo.test

Note that if you apply new changes to the deployment during the canary analysis, Flagger will restart the analysis.

You can monitor all canaries with:

watch kubectl get canaries --all-namespaces
NAMESPACE NAME STATUS WEIGHT LASTTRANSITIONTIME
test podinfo Progressing 15 2019-05-06T14:05:07Z
prod frontend Succeeded 0 2019-05-05T16:15:07Z
prod backend Failed 0 2019-05-04T17:05:07Z

Automated rollback

During the canary analysis you can generate HTTP 500 errors to test if Flagger pauses and rolls back the faulted version.

Trigger another canary deployment:

kubectl -n test set image deployment/podinfo \
podinfod=stefanprodan/podinfo:2.0.2

Generate HTTP 500 errors:

watch curl http://app.example.com/status/500

When the number of failed checks reaches the canary analysis threshold, the traffic is routed back to the primary, the canary is scaled to zero and the rollout is marked as failed.

kubectl -n test describe canary/podinfo
Status:
Canary Weight: 0
Failed Checks: 10
Phase: Failed
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Synced 3m flagger Starting canary deployment for podinfo.test
Normal Synced 3m flagger Advance podinfo.test canary weight 5
Normal Synced 3m flagger Advance podinfo.test canary weight 10
Normal Synced 3m flagger Advance podinfo.test canary weight 15
Normal Synced 3m flagger Halt podinfo.test advancement success rate 69.17% < 99%
Normal Synced 2m flagger Halt podinfo.test advancement success rate 61.39% < 99%
Normal Synced 2m flagger Halt podinfo.test advancement success rate 55.06% < 99%
Normal Synced 2m flagger Halt podinfo.test advancement success rate 47.00% < 99%
Normal Synced 2m flagger (combined from similar events): Halt podinfo.test advancement success rate 38.08% < 99%
Warning Synced 1m flagger Rolling back podinfo.test failed checks threshold reached 10
Warning Synced 1m flagger Canary failed! Scaling down podinfo.test

Custom metrics

The canary analysis can be extended with Prometheus queries.

The demo app is instrumented with Prometheus so you can create a custom check that will use the HTTP request duration histogram to validate the canary.

Edit the canary analysis and add the following metric:

canaryAnalysis:
metrics:
- name: "latency"
threshold: 0.5
interval: 1m
query: |
histogram_quantile(0.99,
sum(
rate(
http_request_duration_seconds_bucket{
kubernetes_namespace="test",
kubernetes_pod_name=~"podinfo-[0-9a-zA-Z]+(-[0-9a-zA-Z]+)"
}[1m]
)
) by (le)
)

The threshold is set to 500ms so if the average request duration in the last minute goes over half a second then the analysis will fail and the canary will not be promoted.

Trigger a canary deployment by updating the container image:

kubectl -n test set image deployment/podinfo \
podinfod=stefanprodan/podinfo:2.0.3

Generate high response latency:

watch curl http://app.exmaple.com/delay/2

Watch Flagger logs:

kubectl -n nginx-ingress logs deployment/flagger -f | jq .msg
Starting canary deployment for podinfo.test
Advance podinfo.test canary weight 5
Advance podinfo.test canary weight 10
Advance podinfo.test canary weight 15
Halt podinfo.test advancement latency 1.20 > 0.5
Halt podinfo.test advancement latency 1.45 > 0.5
Halt podinfo.test advancement latency 1.60 > 0.5
Halt podinfo.test advancement latency 1.69 > 0.5
Halt podinfo.test advancement latency 1.70 > 0.5
Rolling back podinfo.test failed checks threshold reached 5
Canary failed! Scaling down podinfo.test

If you have Slack configured, Flagger will send a notification with the reason why the canary failed.

A/B Testing

Besides weighted routing, Flagger can be configured to route traffic to the canary based on HTTP match conditions. In an A/B testing scenario, you'll be using HTTP headers or cookies to target a certain segment of your users. This is particularly useful for frontend applications that require session affinity.

Flagger A/B Testing Stages

Edit the canary analysis, remove the max/step weight and add the match conditions and iterations:

canaryAnalysis:
interval: 1m
threshold: 10
iterations: 10
match:
# curl -H 'X-Canary: insider' http://app.example.com
- headers:
x-canary:
exact: "insider"
# curl -b 'canary=always' http://app.example.com
- headers:
cookie:
exact: "canary"
metrics:
- name: request-success-rate
threshold: 99
interval: 1m
webhooks:
- name: load-test
url: http://localhost:8888/
timeout: 5s
metadata:
type: cmd
cmd: "hey -z 1m -q 10 -c 2 -H 'Cookie: canary=always' http://app.example.com/"
logCmdOutput: "true"

The above configuration will run an analysis for ten minutes targeting users that have a canary cookie set to always or those that call the service using the X-Canary: insider header.

Trigger a canary deployment by updating the container image:

kubectl -n test set image deployment/podinfo \
podinfod=stefanprodan/podinfo:2.0.4

Flagger detects that the deployment revision changed and starts the A/B testing:

kubectl -n test describe canary/podinfo
Status:
Failed Checks: 0
Phase: Succeeded
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Synced 3m flagger New revision detected podinfo.test
Normal Synced 3m flagger Scaling up podinfo.test
Warning Synced 3m flagger Waiting for podinfo.test rollout to finish: 0 of 1 updated replicas are available
Normal Synced 3m flagger Advance podinfo.test canary iteration 1/10
Normal Synced 3m flagger Advance podinfo.test canary iteration 2/10
Normal Synced 3m flagger Advance podinfo.test canary iteration 3/10
Normal Synced 2m flagger Advance podinfo.test canary iteration 4/10
Normal Synced 2m flagger Advance podinfo.test canary iteration 5/10
Normal Synced 1m flagger Advance podinfo.test canary iteration 6/10
Normal Synced 1m flagger Advance podinfo.test canary iteration 7/10
Normal Synced 55s flagger Advance podinfo.test canary iteration 8/10
Normal Synced 45s flagger Advance podinfo.test canary iteration 9/10
Normal Synced 35s flagger Advance podinfo.test canary iteration 10/10
Normal Synced 25s flagger Copying podinfo.test template spec to podinfo-primary.test
Warning Synced 15s flagger Waiting for podinfo-primary.test rollout to finish: 1 of 2 updated replicas are available
Normal Synced 5s flagger Promotion completed! Scaling down podinfo.test