Canary Deployment Strategy

Introduction

A canary deployment consists of routing a subset of traffic from the old version to the new version of the application based on weight to reduce the risk of deploying a new version that impacts the workload. For example, 90 percent of the traffic goes to the old version, and 10 percent goes to the new version.

This deployment involves making small staged releases and sending them to a small subset of users for testing and observation. Once no bugs are encountered and canary deployment works well, users can easily revert to a new version and delete the old version of the application.

Procedures

Use the following steps to upgrade the SPK Application Pods to a newer version using the Canary deployment.

  1. Install the SPK TCPTestApp with version 0.11.0 using Helm --set option:

    helm install spk-tcp --namespace=xyz-app --version=0.11.0 f5ing-testapp-v1 f5-ingress-dev/f5ing-testapp --set service.port=8051 --set service.protocol=TCP --set app.port=8051 --set service.targetPort=8050 --set ingress.enabled=false --set app.destinationAddress=10.33.0.55 --set app.persist.mode="PERSIST_TYPE_NONE" --set app.persist.timeout=300 --set app.persist.hashAlg="PERSIST_HASH_DEFAULT" --set app.serviceDownAction="POOLMBR_ACTION_REJECT"

    _images/spk_warn.png Important: The SPK Controller watches this app namespace and pushes it to TMM’s load balancing pool member list.

    _images/spk_info.png Note: Once the SPK TCPTestApp is installed, automatically the deployment with version 0.11.0 and the pod under it get created. And also, the ClusterIP service is created.

    • Download to view example status for deployment (deployment_v1 file here) and service (service file here)

  2. Verify the STATUS of the SPK TCPTestApp Pod with version 0.11.0:

    oc get pod -n xyz-app 

    NAME                                READY        STATUS       RESTARTS   AGE
f5ing-testapp-v1-84d985cd-jvn7      1/1          Running      0          8s

  3. Obtain the IP address of the SPK TCPTestApp Pod:

    In this example, SPK TCPTestApp Pod is installed in the xyz-app Project.

    oc get svc -n xyz-app

    In this example, SPK TCPTestApp CLUSTER-IP address is 172.30.234.85 and the PORT(S) is 8051.

    NAME                 TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
f5ing-testapp-v1     ClusterIP   172.30.234.85   <none>        8051/TCP   22s

  4. Scale the SPK TCPTestApp Pod to 2 more in the deployment_v1:

    oc scale --replicas=3 deployment f5ing-testapp-v1 -n xyz-app

  5. Verify the total pods created in deployment_v1:

    oc get pod -n xyz-app

    NAME 	                            READY        STATUS     RESTARTS    AGE     
f5ing-testapp-v1-84d985-dc4cq       0/1          Running    0           3s	     
f5ing-testapp-v1-84d985-hvvr8       1/1          Running    0           18s     
f5ing-testapp-v1-84d985-srcws       1/1          Running    0           18s

  6. Verify the deployment_v1 status:

    oc get deploy -n xyz-app

    NAME                   READY     UP-TO-DATE     AVAILABLE     AGE
f5ing-test-app-v1      3/3       3              3             30s

  7. Create a deployment of version 0.11.4 named deployment_v2.yaml with 1 pod using the download the file here.

    oc create -f deployment_v2.yaml

    _images/spk_warn.png Important: Ensure that the labels in the deployment of version 0.11.4 match those in the deployment of version 0.11.0 so that the pod is created as one more endpoint on the same service.

  8. Verify the details of the deployments created:

    oc get deploy -n xyz

    NAME               READY   UP-TO-DATE   AVAILABLE   AGE
f5ing-testapp-v1   3/3     3            3           2m
f5ing-testapp-v2   1/1     1            1           20s

    _images/spk_info.png Note: Please notice that two deployments are displayed under one service since the labels are the same.

  9. Verify the total pods created in both the deployments:

    oc get pod -n xyz-app -o wide

    In this example, 3 pods of version 0.11.0 and 1 pod of version 0.11.4 are created.

    NAME                              READY   STATUS    RESTARTS    AGE       IP              NODE     
f5ing-testapp-v1-84d985cd-7nplz   1/1     Running   0           4h29m     10.128.2.148    worker-0.ocp3.pd.f5net.com   
f5ing-testapp-v1-84d985cd-gddd9   1/1     Running   0           4h29m     10.128.2.147    worker-0.ocp3.pd.f5net.com   
f5ing-testapp-v1-84d985cd-szc4s   1/1     Running   0           4h29m     10.131.0.201    worker-1.ocp3.pd.f5net.com   
f5ing-testapp-v2-7f69d59fd-nn2vw  1/1     Running   0           4h26m     10.128.2.149    worker-0.ocp3.pd.f5net.com

  10. Verify the traffic is flowing to all the pods:

    nc -v 10.33.0.55 8051

    In this example, the app pod with IP 10.128.2.147 responded:

    Connection to 10.33.0.55 8051 port [tcp/*] succeeded!
Hello 
echo from TCP server [10.128.2.147]: Hello

    In this example, the app pod with IP 10.128.2.148 responded:

    Connection to 10.33.0.55 8051 port [tcp/*] succeeded!
Hi
echo from TCP server [10.128.2.148]: Hi

    In this example, the app pod with IP 10.131.0.201 responded:

    Connection to 10.33.0.55 8051 port [tcp/*] succeeded!
Good Morning 
echo from TCP server [10.131.0.201]: Good Morning

    In this example, the app pod with IP 10.128.2.149 responded

    Connection to 10.33.0.55 8051 port [tcp/*] succeeded!
Good Evening
echo from TCP server [10.128.2.149]: Good Evening

In this way, traffic will be distributed accordingly.

Feedback

Provide feedback to improve this document by emailing spkdocs@f5.com.