stable
Install
Tutorials
Linkerd Canary Deployments
This guide shows you how to use Linkerd and Flagger to automate canary deployments.
Flagger Linkerd Traffic Split
Prerequisites
Flagger requires a Kubernetes cluster v1.16 or newer and Linkerd 2.10 or newer.
Install Linkerd the Promethues (part of Linkerd Viz):
1
linkerd install | kubectl apply -f -
2
linkerd viz install | kubectl apply -f -
Copied!
Install Flagger in the linkerd namespace:
1
kubectl apply -k github.com/fluxcd/flagger//kustomize/linkerd
Copied!
Bootstrap
Flagger takes a Kubernetes deployment and optionally a horizontal pod autoscaler (HPA), then creates a series of objects (Kubernetes deployments, ClusterIP services and SMI traffic split). These objects expose the application inside the mesh and drive the canary analysis and promotion.
Create a test namespace and enable Linkerd proxy injection:
1
kubectl create ns test
2
kubectl annotate namespace test linkerd.io/inject=enabled
Copied!
Install the load testing service to generate traffic during the canary analysis:
1
kubectl apply -k https://github.com/fluxcd/flagger//kustomize/tester?ref=main
Copied!
Create a deployment and a horizontal pod autoscaler:
1
kubectl apply -k https://github.com/fluxcd/flagger//kustomize/podinfo?ref=main
Copied!
Create a canary custom resource for the podinfo deployment:
1
apiVersion: flagger.app/v1beta1
2
kind: Canary
3
metadata:
4
name: podinfo
5
namespace: test
6
spec:
7
# deployment reference
8
targetRef:
9
apiVersion: apps/v1
10
kind: Deployment
11
name: podinfo
12
# HPA reference (optional)
13
autoscalerRef:
14
apiVersion: autoscaling/v2beta2
15
kind: HorizontalPodAutoscaler
16
name: podinfo
17
# the maximum time in seconds for the canary deployment
18
# to make progress before it is rollback (default 600s)
19
progressDeadlineSeconds: 60
20
service:
21
# ClusterIP port number
22
port: 9898
23
# container port number or name (optional)
24
targetPort: 9898
25
analysis:
26
# schedule interval (default 60s)
27
interval: 30s
28
# max number of failed metric checks before rollback
29
threshold: 5
30
# max traffic percentage routed to canary
31
# percentage (0-100)
32
maxWeight: 50
33
# canary increment step
34
# percentage (0-100)
35
stepWeight: 5
36
# Linkerd Prometheus checks
37
metrics:
38
- name: request-success-rate
39
# minimum req success rate (non 5xx responses)
40
# percentage (0-100)
41
thresholdRange:
42
min: 99
43
interval: 1m
44
- name: request-duration
45
# maximum req duration P99
46
# milliseconds
47
thresholdRange:
48
max: 500
49
interval: 30s
50
# testing (optional)
51
webhooks:
52
- name: acceptance-test
53
type: pre-rollout
54
url: http://flagger-loadtester.test/
55
timeout: 30s
56
metadata:
57
type: bash
58
cmd: "curl -sd 'test' http://podinfo-canary.test:9898/token | grep token"
59
- name: load-test
60
type: rollout
61
url: http://flagger-loadtester.test/
62
metadata:
63
cmd: "hey -z 2m -q 10 -c 2 http://podinfo-canary.test:9898/"
Copied!
Save the above resource as podinfo-canary.yaml and then apply it:
1
kubectl apply -f ./podinfo-canary.yaml
Copied!
When the canary analysis starts, Flagger will call the pre-rollout webhooks before routing traffic to the canary. The canary analysis will run for five minutes while validating the HTTP metrics and rollout hooks every half a minute.
After a couple of seconds Flagger will create the canary objects:
1
# applied
2
deployment.apps/podinfo
3
horizontalpodautoscaler.autoscaling/podinfo
4
ingresses.extensions/podinfo
5
canary.flagger.app/podinfo
6
​
7
# generated
8
deployment.apps/podinfo-primary
9
horizontalpodautoscaler.autoscaling/podinfo-primary
10
service/podinfo
11
service/podinfo-canary
12
service/podinfo-primary
13
trafficsplits.split.smi-spec.io/podinfo
Copied!
After the boostrap, the podinfo deployment will be scaled to zero and the traffic to
podinfo.test will be routed to the primary pods. During the canary analysis, the podinfo-canary.test address can be used to target directly the canary pods.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:
1
kubectl -n test set image deployment/podinfo \
2
podinfod=ghcr.io/stefanprodan/podinfo:6.0.1
Copied!
Flagger detects that the deployment revision changed and starts a new rollout:
1
kubectl -n test describe canary/podinfo
2
​
3
Status:
4
Canary Weight: 0
5
Failed Checks: 0
6
Phase: Succeeded
7
Events:
8
New revision detected! Scaling up podinfo.test
9
Waiting for podinfo.test rollout to finish: 0 of 1 updated replicas are available
10
Pre-rollout check acceptance-test passed
11
Advance podinfo.test canary weight 5
12
Advance podinfo.test canary weight 10
13
Advance podinfo.test canary weight 15
14
Advance podinfo.test canary weight 20
15
Advance podinfo.test canary weight 25
16
Waiting for podinfo.test rollout to finish: 1 of 2 updated replicas are available
17
Advance podinfo.test canary weight 30
18
Advance podinfo.test canary weight 35
19
Advance podinfo.test canary weight 40
20
Advance podinfo.test canary weight 45
21
Advance podinfo.test canary weight 50
22
Copying podinfo.test template spec to podinfo-primary.test
23
Waiting for podinfo-primary.test rollout to finish: 1 of 2 updated replicas are available
24
Promotion completed! Scaling down podinfo.test
Copied!
Note that if you apply new changes to the deployment during the canary analysis, Flagger will restart the analysis.
A canary deployment is triggered by changes in any of the following objects:
- Deployment PodSpec (container image, command, ports, env, resources, etc)
- ConfigMaps mounted as volumes or mapped to environment variables
- Secrets mounted as volumes or mapped to environment variables
You can monitor all canaries with:
1
watch kubectl get canaries --all-namespaces
2
​
3
NAMESPACE NAME STATUS WEIGHT LASTTRANSITIONTIME
4
test podinfo Progressing 15 2019-06-30T14:05:07Z
5
prod frontend Succeeded 0 2019-06-30T16:15:07Z
6
prod backend Failed 0 2019-06-30T17:05:07Z
Copied!
Automated rollback
During the canary analysis you can generate HTTP 500 errors and high latency to test if Flagger pauses and rolls back the faulted version.
Trigger another canary deployment:
1
kubectl -n test set image deployment/podinfo \
2
podinfod=ghcr.io/stefanprodan/podinfo:6.0.2
Copied!
Exec into the load tester pod with:
1
kubectl -n test exec -it flagger-loadtester-xx-xx sh
Copied!
Generate HTTP 500 errors:
1
watch -n 1 curl http://podinfo-canary.test:9898/status/500
Copied!
Generate latency:
1
watch -n 1 curl http://podinfo-canary.test:9898/delay/1
Copied!
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.
1
kubectl -n test describe canary/podinfo
2
​
3
Status:
4
Canary Weight: 0
5
Failed Checks: 10
6
Phase: Failed
7
Events:
8
Starting canary analysis for podinfo.test
9
Pre-rollout check acceptance-test passed
10
Advance podinfo.test canary weight 5
11
Advance podinfo.test canary weight 10
12
Advance podinfo.test canary weight 15
13
Halt podinfo.test advancement success rate 69.17% < 99%
14
Halt podinfo.test advancement success rate 61.39% < 99%
15
Halt podinfo.test advancement success rate 55.06% < 99%
16
Halt podinfo.test advancement request duration 1.20s > 0.5s
17
Halt podinfo.test advancement request duration 1.45s > 0.5s
18
Rolling back podinfo.test failed checks threshold reached 5
19
Canary failed! Scaling down podinfo.test
Copied!
Custom metrics
The canary analysis can be extended with Prometheus queries.
Let's a define a check for not found errors. Edit the canary analysis and add the following metric:
1
analysis:
2
metrics:
3
- name: "404s percentage"
4
threshold: 3
5
query: |
6
100 - sum(
7
rate(
8
response_total{
9
namespace="test",
10
deployment="podinfo",
11
status_code!="404",
12
direction="inbound"
13
}[1m]
14
)
15
)
16
/
17
sum(
18
rate(
19
response_total{
20
namespace="test",
21
deployment="podinfo",
22
direction="inbound"
23
}[1m]
24
)
25
)
26
* 100
Copied!
The above configuration validates the canary version by checking if the HTTP 404 req/sec percentage is below three percent of the total traffic. If the 404s rate reaches the 3% threshold, then the analysis is aborted and the canary is marked as failed.
Trigger a canary deployment by updating the container image:
1
kubectl -n test set image deployment/podinfo \
2
podinfod=ghcr.io/stefanprodan/podinfo:6.0.3
Copied!
Generate 404s:
1
watch -n 1 curl http://podinfo-canary:9898/status/404
Copied!
Watch Flagger logs:
1
kubectl -n linkerd logs deployment/flagger -f | jq .msg
2
​
3
Starting canary deployment for podinfo.test
4
Pre-rollout check acceptance-test passed
5
Advance podinfo.test canary weight 5
6
Halt podinfo.test advancement 404s percentage 6.20 > 3
7
Halt podinfo.test advancement 404s percentage 6.45 > 3
8
Halt podinfo.test advancement 404s percentage 7.22 > 3
9
Halt podinfo.test advancement 404s percentage 6.50 > 3
10
Halt podinfo.test advancement 404s percentage 6.34 > 3
11
Rolling back podinfo.test failed checks threshold reached 5
12
Canary failed! Scaling down podinfo.test
Copied!
If you have Slack configured, Flagger will send a notification with the reason why the canary failed.
Linkerd Ingress
There are two ingress controllers that are compatible with both Flagger and Linkerd: NGINX and Gloo.
Install NGINX:
1
helm upgrade -i nginx-ingress stable/nginx-ingress \
2
--namespace ingress-nginx
Copied!
Create an ingress definition for podinfo that rewrites the incoming header to the internal service name (required by Linkerd):
1
apiVersion: extensions/v1beta1
2
kind: Ingress
3
metadata:
4
name: podinfo
5
namespace: test
6
labels:
7
app: podinfo
8
annotations:
9
kubernetes.io/ingress.class: "nginx"
10
nginx.ingress.kubernetes.io/configuration-snippet: |
11
proxy_set_header l5d-dst-override $service_name.$namespace.svc.cluster.local:9898;
12
proxy_hide_header l5d-remote-ip;
13
proxy_hide_header l5d-server-id;
14
spec:
15
rules:
16
- host: app.example.com
17
http:
18
paths:
19
- backend:
20
serviceName: podinfo
21
servicePort: 9898
Copied!
When using an ingress controller, the Linkerd traffic split does not apply to incoming traffic since NGINX in running outside of the mesh. In order to run a canary analysis for a frontend app, Flagger creates a shadow ingress and sets the NGINX specific annotations.
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 Linkerd Ingress
Edit podinfo canary analysis, set the provider to
nginx, add the ingress reference, remove the max/step weight and add the match conditions and iterations:1
apiVersion: flagger.app/v1beta1
2
kind: Canary
3
metadata:
4
name: podinfo
5
namespace: test
6
spec:
7
# ingress reference
8
provider: nginx
9
ingressRef:
10
apiVersion: extensions/v1beta1
11
kind: Ingress
12
name: podinfo
13
targetRef:
14
apiVersion: apps/v1
15
kind: Deployment
16
name: podinfo
17
autoscalerRef:
18
apiVersion: autoscaling/v2beta2
19
kind: HorizontalPodAutoscaler
20
name: podinfo
21
service:
22
# container port
23
port: 9898
24
analysis:
25
interval: 1m
26
threshold: 10
27
iterations: 10
28
match:
29
# curl -H 'X-Canary: always' http://app.example.com
30
- headers:
31
x-canary:
32
exact: "always"
33
# curl -b 'canary=always' http://app.example.com
34
- headers:
35
cookie:
36
exact: "canary"
37
# Linkerd Prometheus checks
38
metrics:
39
- name: request-success-rate
40
thresholdRange:
41
min: 99
42
interval: 1m
43
- name: request-duration
44
thresholdRange:
45
max: 500
46
interval: 30s
47
webhooks:
48
- name: acceptance-test
49
type: pre-rollout
50
url: http://flagger-loadtester.test/
51
timeout: 30s
52
metadata:
53
type: bash
54
cmd: "curl -sd 'test' http://podinfo-canary:9898/token | grep token"
55
- name: load-test
56
type: rollout
57
url: http://flagger-loadtester.test/
58
metadata:
59
cmd: "hey -z 2m -q 10 -c 2 -H 'Cookie: canary=always' http://app.example.com"
Copied!
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: always header.Note that the load test now targets the external address and uses the canary cookie.
Trigger a canary deployment by updating the container image:
1
kubectl -n test set image deployment/podinfo \
2
podinfod=ghcr.io/stefanprodan/podinfo:6.0.4
Copied!
Flagger detects that the deployment revision changed and starts the A/B testing:
1
kubectl -n test describe canary/podinfo
2
​
3
Events:
4
Starting canary deployment for podinfo.test
5
Pre-rollout check acceptance-test passed
6
Advance podinfo.test canary iteration 1/10
7
Advance podinfo.test canary iteration 2/10
8
Advance podinfo.test canary iteration 3/10
9
Advance podinfo.test canary iteration 4/10
10
Advance podinfo.test canary iteration 5/10
11
Advance podinfo.test canary iteration 6/10
12
Advance podinfo.test canary iteration 7/10
13
Advance podinfo.test canary iteration 8/10
14
Advance podinfo.test canary iteration 9/10
15
Advance podinfo.test canary iteration 10/10
16
Copying podinfo.test template spec to podinfo-primary.test
17
Waiting for podinfo-primary.test rollout to finish: 1 of 2 updated replicas are available
18
Promotion completed! Scaling down podinfo.test
Copied!
The above procedure can be extended with custom metrics checks, webhooks, manual promotion approval and Slack or MS Teams notifications.
Last modified 2mo ago