Container Service for Kubernetes (ACK) allows you to mirror network traffic across
clusters. Traffic mirroring is used in system simulations and troubleshooting. This
topic describes how to use an Ingress controller to mirror network traffic.
Prerequisites
Two ACK clusters are created. For more information, see
Create an ACK managed cluster. In this topic, a cluster named K8s Product Cluster is used as the production environment
and another cluster named K8s Stage Cluster is used as the staging environment.
Scenarios
Traffic mirroring can be used in the following scenarios:
- Before you make significant changes to a system or release new features, you must
run stress tests to evaluate the stability of the system. In most cases, production
workloads are simulated in a staging environment to test the stability of a new system
before the system is released. However, the actual loads are difficult to estimate
because the system may receive both normal and abnormal network traffic. To resolve
this issue, you can mirror network traffic from applications that are deployed in
the production environment to the staging environment. Then, you can simulate the
production workloads in the staging environment.
- When a system deployed in the production environment encounters a performance bottleneck
and you cannot locate the cause, you can mirror the network traffic of the system
to a staging environment. This way, you can troubleshoot errors in the staging environment.
Step 1: Deploy basic applications
- Deploy applications in the cluster named K8s Product Cluster and test access to the
applications.
- Deploy an application in the cluster named K8s Product Cluster and use an Ingress
to expose the Services in the cluster.
apiVersion: apps/v1
kind: Deployment
metadata:
name: nginx-deployment
spec:
replicas: 1
selector:
matchLabels:
app: nginx
template:
metadata:
labels:
app: nginx
spec:
containers:
# A previous version of the image is deployed.
- image: registry.cn-hangzhou.aliyuncs.com/acs-sample/old-nginx
imagePullPolicy: Always
name: nginx
ports:
- containerPort: 80
protocol: TCP
restartPolicy: Always
---
apiVersion: v1
kind: Service
metadata:
name: nginx-service
spec:
ports:
- port: 80
protocol: TCP
targetPort: 80
selector:
app: nginx
type: NodePort
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: nginx-ingress
spec:
rules:
# By default, the application is deployed in the region where the cluster resides. You can specify a region and configure DNS resolution.
- host: nginx.c37bf6b77bded43669ba2fb67448b****.cn-hangzhou.alicontainer.com
http:
paths:
- path: /
backend:
service:
name: nginx-service
port:
number: 80
pathType: ImplementationSpecific
- Run the following command to view the Ingress configurations:
kubectl get ing nginx-ingress
Expected output:
NAME HOSTS ADDRESS PORTS AGE
nginx-ingress nginx.c37bf6b77bded43669ba2fb67448b****.cn-hangzhou.alicontainer.com 47.110.1**.** 80 8m
- Run the following command to check whether you can access the domain name of the application:
curl http://nginx.c37bf6b77bded43669ba2fb67448b****.cn-hangzhou.alicontainer.com
- Deploy the same application in the cluster named K8s Stage Cluster and test access
to the application.
- Deploy the same application in the cluster named K8s Stage Cluster and use an Ingress
to expose the Services in the cluster.
apiVersion: apps/v1
kind: Deployment
metadata:
name: nginx-deployment
spec:
replicas: 1
selector:
matchLabels:
app: nginx
template:
metadata:
labels:
app: nginx
spec:
containers:
# The latest version of the image is deployed.
- image: registry.cn-hangzhou.aliyuncs.com/acs-sample/new-nginx
imagePullPolicy: Always
name: nginx
ports:
- containerPort: 80
protocol: TCP
restartPolicy: Always
---
apiVersion: v1
kind: Service
metadata:
name: nginx-service
spec:
ports:
- port: 80
protocol: TCP
targetPort: 80
selector:
app: nginx
type: NodePort
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: nginx-ingress
spec:
rules:
# By default, the application is deployed in the region where the cluster resides. You can specify a region and configure DNS resolution.
- host: nginx.c41eb6ca34a3e49f7aea63b8bc9e8****.cn-beijing.alicontainer.com
http:
paths:
- path: /
backend:
service:
name: nginx-service
port:
number: 80
pathType: ImplementationSpecific
- Run the following command to view the Ingress configurations:
kubectl get ing nginx-ingress
Expected output:
NAME HOSTS ADDRESS PORTS AGE
nginx-ingress nginx.c41eb6ca34a3e49f7aea63b8bc9e8****.cn-beijing.alicontainer.com 39.106.***.* 80 1m
- Run the following command to check whether you can access the domain name of the application:
curl http://nginx.c41eb6ca34a3e49f7aea63b8bc9e8****.cn-beijing.alicontainer.com
Step 2: Mirror the network traffic
Mirror all network traffic of the application deployed in the cluster K8s Product
Cluster to Services in the cluster K8s Stage Cluster. This way, all requests that
are sent to the domain name
nginx.c37bf6b77bded43669ba2fb67448b****.cn-hangzhou.alicontainer.com are replicated to the domain name
nginx.c41eb6ca34a3e49f7aea63b8bc9e8****.cn-beijing.alicontainer.com.
- Run the following command to modify nginx-configuration. Set the percentage of network traffic that you want to mirror and specify the domain
name that receives the mirrored network traffic:
kubectl -n kube-system edit cm nginx-configuration
Add the following content to nginx-configuration:
http-snippet: |
split_clients "$date_gmt" $mirror_servers {
100% nginx.c41eb6ca34a3e49f7aea63b8bc9e8****.cn-beijing.alicontainer.com;
}
# Notes on configurations:
# 1. The percentage of network traffic that you want to mirror must be greater than 0 and can be up to 100. The sum of percentage cannot exceed 100.
# 2. You can set one or more applications to receive the mirrored network traffic.
- Run the following command to modify nginx-ingress:
You can add traffic mirroring settings to the source Ingress by setting the configuration-snippet
and server-snippet annotations.
kubectl edit ingress nginx-ingress
Modify configuration-snippet and server-snippet of the nginx-ingress controller based on your business requirements. The following
YAML file is an example:
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: nginx-ingress
annotations:
nginx.ingress.kubernetes.io/configuration-snippet: |
mirror /mirror; # You can repeat this setting N times to mirror the network traffic N times.
nginx.ingress.kubernetes.io/server-snippet: |
location = /mirror {
internal;
# Does not print the log of mirrored requests
#access_log off;
# Set proxy_upstream_name in the following format: [Namespace]-[BackendServiceName]-[BackendServicePort]
set $proxy_upstream_name "default-nginx-service-80";
# A custom string that is carried in the request header X-Shadow-Service and then passed to the mirror server.
set $shadow_service_name "nginx-product-service";
proxy_set_header X-Shadow-Service $shadow_service_name;
proxy_set_header Host $mirror_servers;
proxy_pass http://$mirror_servers$request_uri;
}
spec:
rules:
- host: nginx.c37bf6b77bded43669ba2fb67448b****.cn-hangzhou.alicontainer.com
http:
paths:
- path: /
backend:
service:
name: nginx-service
port:
number: 80
pathType: ImplementationSpecific
Verify the result.
Access the domain name nginx.c37bf6b77bded43669ba2fb67448b****.cn-hangzhou.alicontainer.com. This domain name belongs to the application that is deployed in the cluster K8s
Product Cluster. Run the following command. The output shows that each request sent
to the domain name of the application deployed in the cluster K8s Product Cluster
is replicated, and then sent to the application deployed in the cluster K8s Stage
Cluster.
kubectl -n kube-system logs --tail=0 -f nginx-ingress-controller-674c96ffbc-9mc8n
