DEV Community

Arseny Zinchenko
Arseny Zinchenko

Posted on • Originally published at rtfm.co.ua on

Istio: shared Ingress/AWS ALB, Helm chart with conditions, Istio, and ExternalDNS

Let’s proceed with our journey with Istio.

Previous parts:

  1. Istio: an overview and running Service Mesh in Kubernetes
  2. Istio: external AWS Application LoadBalancer and Istio Ingress Gateway

Besides Istio, in this post, we will also configure ExternalDNS, see the Kubernetes: update AWS Route53 DNS from an Ingress for details.

Everything described below is a kind of Proof of Concept and will be deployed to the same AWS Elastic Kubernetes Service Dev cluster.

Goals

So, now I have three tasks:

  1. test how will work one shared AWS Application LoadBalancer (AWS ALB) and Istio Ingress Gateway with applications in different namespaces
  2. create a Helm chart with templates to be able to select to create an Ingress, Istio Gateway, and Istio VirtualService
  3. configure ExternalDNS to create records in the AWS Route53 when adding an Istio Gateway or VirtualService

At first, let’s see how Istio Ingress Gateway will work with applications, located in dedicated namespaces. For this, we will create an Ingress which will create an AWS Application LoadBalancer with ALB Ingress Controller, and two testing applications, each with its own Service, Gateway, and VirtualService.

In the Ingress/AWS ALB we will describe hosts and this will trigger ExternalDNS to create records. Also, here we will do SSL termination — will attach SSL certificates from the AWS Certificate Manager.

In the Gateway of each of those applications, we will open ports on the Istio Ingress Gateway, and add hosts for which this Gateway will accept traffic.

Shared Ingress will be created in the istio-system namespace as it needs to access Istio Ingress Service.

Actually, we could create one shared Gateway, but at the end of the discussion here>>> people saying that will be better to have a dedicated Gateway per each application, and this seems to be more correct, so let’s do it in this way.

Thus, now we will create:

  1. dedicated Ingress/AWS ALB with two test records in Route53 in the istio-system namespace
  2. test app1 and app2 in namespaces backend-app-1-ns and backend-app-2-ns accordingly, each app with its own Deployment, Service, Gateway, and VirtualService

The second task will more interesting: will need to create Helm templates to deploy applications on different environments (kind of Dev and Prod) to deploy applications on various environments with different Ingress and Istio configurations.

And in the third task, we will configure ExternalDNS to work with Istio Gateway and VirtualService.

Some parts below could be confusing but I’ve tried to describe them as simply as I know-how.

Shared Ingress/AWS ALB

Create a manifest file for the Ingress/ALB — common-ingress-gateway.yaml:

--------
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: backend-common-alb
  namespace: istio-system
  annotations:
    # create AWS Application LoadBalancer
    kubernetes.io/ingress.class: alb
    # external type
    alb.ingress.kubernetes.io/scheme: internet-facing
    # AWS Certificate Manager certificate's ARN
    alb.ingress.kubernetes.io/certificate-arn: "arn:aws:acm:us-east-2:534***385:certificate/db886018-c173-43d0-b1be-b940de71f2a2"
    # open ports 80 and 443 
    alb.ingress.kubernetes.io/listen-ports: '[{"HTTP": 80}, {"HTTPS":443}]'
    # redirect all HTTP to HTTPS
    alb.ingress.kubernetes.io/actions.ssl-redirect: '{"Type": "redirect", "RedirectConfig": { "Protocol": "HTTPS", "Port": "443", "StatusCode": "HTTP_301"}}'
    # ExternalDNS settings: [https://rtfm.co.ua/en/kubernetes-update-aws-route53-dns-from-an-ingress/](https://rtfm.co.ua/en/kubernetes-update-aws-route53-dns-from-an-ingress/)
    external-dns.alpha.kubernetes.io/hostname: "app1-test-common-ingress.example.com, app2-test-common-ingress.example.com"
spec:
  rules:
  - http:
      paths:
        - path: /*
          backend:
            serviceName: ssl-redirect
            servicePort: use-annotation
        - path: /*
          backend:
            serviceName: istio-ingressgateway
            servicePort: 80
Enter fullscreen mode Exit fullscreen mode

Here, in the annotations, we are describing domains to be created by the ExternalDNS, and they will be mapped to the URL of the AWS ALB, and in the backend, we will send all traffic to the istio-ingressgateway Kubernetes Service.

Deploy it:

$ kubectl apply -f common-ingress-gateway.yaml
ingress.extensions/backend-common-alb created
Enter fullscreen mode Exit fullscreen mode

Check the Ingress/ALB:

$ kubectl -n istio-system get ingress backend-common-alb
NAME CLASS HOSTS ADDRESS PORTS AGE
backend-common-alb <none> * aadca942-***1826302788.us-east-2.elb.amazonaws.com 80 72s
Enter fullscreen mode Exit fullscreen mode

And ExternalDNS logs:

…
time=”2021–04–12T09:45:02Z” level=info msg=”Desired change: CREATE app-1-test-common-ingress.example.com A [Id: /hostedzone/Z30KLN6M3D0LB6]”
time=”2021–04–12T09:45:02Z” level=info msg=”Desired change: CREATE app-1-test-common-ingress.example.com TXT [Id: /hostedzone/Z30KLN6M3D0LB6]”
time=”2021–04–12T09:45:02Z” level=info msg=”Desired change: CREATE app-2-test-common-ingress.example.com A [Id: /hostedzone/Z30KLN6M3D0LB6]”
time=”2021–04–12T09:45:02Z” level=info msg=”Desired change: CREATE app-2-test-common-ingress.example.com TXT [Id: /hostedzone/Z30KLN6M3D0LB6]”
time=”2021–04–12T09:45:03Z” level=info msg=”4 record(s) in zone example.com. [Id: /hostedzone/Z30KLN6M3D0LB6] were successfully updated”
…
Enter fullscreen mode Exit fullscreen mode

Try access the URL, must get an error as Istio Ingress Gateway is not configured yet:

$ curl -I [https://app-1-test-common-ingress.example.com](https://app-1-test-common-ingress.example.com)
HTTP/2 502
date: Mon, 12 Apr 2021 09:46:11 GMT
server: istio-envoy
Enter fullscreen mode Exit fullscreen mode

Yup, the 502 error as Istio Ingress Gateway has no routes for the domains yet:

$ istioctl proxy-config routes -n istio-system istio-ingressgateway-8459df68cb-bh76b
NOTE: This output only contains routes loaded via RDS.
NAME DOMAINS MATCH VIRTUAL SERVICE
* /healthz/ready*
* /stats/prometheus*
Enter fullscreen mode Exit fullscreen mode

Now, let’s create testing applications where we will describe settings for the Istio Ingress Gateway.

Testing applications

Both applications are absolutely similar excepting namespaces and names of applications and services.

Here, we will create:

  • Namespace
  • Deployment
  • Service
  • Gateway
  • VirtualService

The whole manifest is the next:

--------
apiVersion: v1
kind: Namespace
metadata:
  name: backend-app-1-ns
  labels:
    istio-injection:
      enabled
--------
apiVersion: apps/v1
kind: Deployment
metadata:
  name: backend-app-1-deploy
  namespace: backend-app-1-ns
  labels:
    app: backend-app-1
    version: v1
spec:
  replicas: 2
  selector:
    matchLabels:
      app: backend-app-1
  template:
    metadata:
      labels:
        app: backend-app-1
        version: v1
    spec: 
      containers:
      - name: app1
        image: nginxdemos/hello
        ports:
        - containerPort: 80
        resources:
          requests:
            cpu: 100m
            memory: 100Mi
        readinessProbe:
          httpGet:
            path: /
            port: 80
--------
apiVersion: v1
kind: Service
metadata:
  name: backend-app-1-servcie
  namespace: backend-app-1-ns
spec:       
  selector:
    app: backend-app-1
  ports:    
    - name: http
      protocol: TCP
      port: 80
--------
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: backend-app-1-gateway
  namespace: backend-app-1-ns
spec:
  selector:
    istio: ingressgateway
  servers:
  - port:
      number: 80
      name: http
      protocol: HTTP
    hosts:
    - "app-1-test-common-ingress.example.com"
--------
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: backend-app-1-virtualservice
  namespace: backend-app-1-ns
spec:
  hosts:
  - "app-1-test-common-ingress.example.com"
  gateways:
  - backend-app-1-gateway
  http:
  - match:
    - uri:
        prefix: /
    route:
    - destination:
        host: backend-app-1-servcie
        port:
          number: 80
Enter fullscreen mode Exit fullscreen mode

Deploy it:

$ kubectl apply -f app1.yaml
namespace/backend-app-1-ns created
deployment.apps/backend-app-1-deploy created
service/backend-app-1-servcie created
gateway.networking.istio.io/backend-app-1-gateway created
virtualservice.networking.istio.io/backend-app-1-virtualservice created
Enter fullscreen mode Exit fullscreen mode

Copy the manifest’s file, change app-1 to app-2, deploy too:

$ kubectl apply -f app2.yaml
namespace/backend-app-2-ns created
deployment.apps/backend-app-2-deploy created
service/backend-app-2-servcie created
gateway.networking.istio.io/backend-app-2-gateway created
virtualservice.networking.istio.io/backend-app-2-virtualservice created
Enter fullscreen mode Exit fullscreen mode

Check Istio’s routes again:

$ istioctl proxy-config routes -n istio-system istio-ingressgateway-8459df68cb-bh76b=
NOTE: This output only contains routes loaded via RDS.
NAME DOMAINS MATCH VIRTUAL SERVICE
http.80 app-1-test-common-ingress.example.com /* backend-app-1-virtualservice.backend-app-1-ns
http.80 app-2-test-common-ingress.example.com /* backend-app-2-virtualservice.backend-app-2-ns
Enter fullscreen mode Exit fullscreen mode

And try to access the URL of the app-1:

$ curl -I [https://app-1-test-common-ingress.example.com](https://app-1-test-common-ingress.example.com)
HTTP/2 200
date: Mon, 12 Apr 2021 09:52:13 GMT
content-type: text/html
server: istio-envoy
Enter fullscreen mode Exit fullscreen mode

Cool, everything is working now. Now, we have one shared AWS ALB created with Kubernetes Ingress which sends traffic via Istio Ingress Gateway to two dedicated applications and dedicated namespaces.

Drop resources, but leave Ingress with the shared LoadBalancer for further tests:

$ kubectl delete -f app1.yaml
$ kubectl delete -f app2.yaml
Enter fullscreen mode Exit fullscreen mode

And let’s go to the Helm chart.

Istio: shared Ingress/AWS ALB, and Helm

Planning: conditions

The following task is non-trivial: we have a Dev EKS cluster, and a Production EKS кластер.

On the Dev cluster, we already have Istio and it’s testing there, but not for all applications running there. On the Production cluster, we didn’t install Ietio yet but will do it in the future.

Also, all our applications now are leaving in dedicated namespaces, and have dedicated Ingress/AWS ALB.

On the Dev cluster I’d like to change this approach, and use now shared Ingress/ALB and send all traffic via Istio Ingress Gateway, but on the Production — leave it as it is, i.e. each application will use its own ALB, and for now, they will send traffic to an application’s Service directly, and in the future, when we’ll implement Istio on Production, we’d need to change Ingress to use Istio Ingress Gateway.

But still, and Dev and Production applications may use or not Istio, as it’s in the “early-stage” yet in our architecture, and some time settings for application will differ, so need to create such a Helm chart, which could configure necessary resources and Ingress settings.

So, at first, need to Т.е. во-первых — надо determine will a chart uses its own Ingress or a shared one? The second thing is if it’s using its own Ingress, which backend will it use — Istio Ingress Gateway or a common Service of an application? And if it will use Istio Ingress Gateway, the chart have to create Gateway and VirtualService resources.

Thus, our template must accept three configuration schemes:

  1. use a shared LoadBalancer and Istio: Istio Ingress Gateway LoadBalancer and Istio Ingress Gateway Service as a backend
  2. own/dedicated LoadBalancer for an application and Istio Ingress Gateway: create an Ingress for an application with the Istio Ingress Gateway Service as backend
  3. own/dedicated LoadBalancer for an application but without Istio: create an Ingress for an application with a Services of the application as a backend

I.e:

  1. an Ingress shared or own?
  2. if shared — do not create an Ingress resource
  3. if own, then create an Ingress but with choice of the backend — Istio Ingress Service, or a common Kubernetes Service of an application
  4. a backend for an Ingress — Istio or a Service of an application?
  5. if Istio, then need to set the backend as serviceName: istio-ingressgateway, and create Gateway and VirtualService resources
  6. if a Service of an application, then set the baсkend as serviceName: <APPNAME>-service, and do not create Gateway and VirtualService resources

To achieve this goal, let’s use a values.yaml - we have dedicated files for Dev and Production environments.

In those files, we can define two parameters  —  istio.enabled and ingress.enabled.

This will give us an ability to set for Dev ingress.enabled=false, and do not create an Ingress, but set istio.enabled==true and create Gateway and VirtualService, that will be used by the shared Ingress/ALB from the istio-system namespace.

And for the Production, we will able to set ingress.enabled=true and istio.enabled=false, then the chart will be deployed by the currently used scheme, and later, when we will implement Istio on the Production cluster, we will set those values as ingress.enabled=true and istio.enabled=true, and this will create a dedicated Ingress/LoadBalancer, which will send traffic via Istio Ingress Gateway.

Well, let’s try.

Creating Helm chart and templates

Create a new chart:

$ helm create app-1
Creating app-1
Enter fullscreen mode Exit fullscreen mode

Create directories to keep values.yaml for Dev and Production:

$ mkdir -p app-1/env/{dev,prod}
Enter fullscreen mode Exit fullscreen mode

Delete default templates and values:

$ rm -r app-1/templates/* app-1/values.yaml
Enter fullscreen mode Exit fullscreen mode

Create own values files in those directories:

$ vim -p app-1/env/dev/values.yaml app-1/env/prod/values.yaml
Enter fullscreen mode Exit fullscreen mode

For Dev  —  app-1/env/dev/values.yaml:

appConfig:
  name: "backend-app-1"
  version: "v1"
  url: "dev-app-1-test-common-ingress.example.com"

istio:
  enabled: true

ingress:
  enabled: false
Enter fullscreen mode Exit fullscreen mode

Do not create an Ingress, but create Gateway and VirtualService.

Production  —  app-1/env/prod/values.yaml:

appConfig:
  name: "backend-app-1"
  version: "v1"
  url: "prod-app-1-test-common-ingress.example.com"

istio:
  enabled: false

ingress:
  enabled: true
Enter fullscreen mode Exit fullscreen mode

Here, an Ingress will be created as a dedicated AWS ALB, but resources for will not be created — the Ingress will send traffic directly to a Service of the application

Create template files:

$ vim -p app-1/templates/ingress.yaml app-1/templates/service.yaml app-1/templates/deployment.yaml
Enter fullscreen mode Exit fullscreen mode

We will not define a namespace in those templates (only for the Ingress, see below), as a namespace will be created by Helm during deployment.

Deployment

No changes in here, just some values are taken from the values.yaml:

-------- 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: {{ .Values.appConfig.name }}-deploy
  labels:
    app: {{ .Values.appConfig.name }}
    version: {{ .Values.appConfig.version }}
spec:
  replicas: 2
  selector:
    matchLabels:
      app: {{ .Values.appConfig.name }}
  template: 
    metadata:
      labels:
        app: {{ .Values.appConfig.name }}
        version: {{ .Values.appConfig.version }}
    spec: 
      containers:
      - name: web-app
        image: nginxdemos/hello
        ports:
        - containerPort: 80
        resources:
          requests:
            cpu: 100m
            memory: 100Mi
        readinessProbe:
          httpGet:
            path: /
            port: 80
Enter fullscreen mode Exit fullscreen mode

Service, VirtualService, Gateway

Here, we will always create a Service with a condition check: if ingress.enabled==true, then set type: NodePort, so our LoadBalancer will be able to send traffic to a WorkerNode. If it's false, then use the default value ClusterIP, so our requests will not go through additional Iptables rules but instead will be sent directly to a WorkerNode where a Pod is living (read also Kubernetes: Service, load balancing, kube-proxy, and iptables):

-------- 
apiVersion: v1
kind: Service
metadata:
  name: {{ .Values.appConfig.name }}-service
spec:
  {{- if .Values.ingress.enabled }}
  type: NodePort
  {{- end }}
  selector:
    app: {{ .Values.appConfig.name }}
  ports:
    - name: http
      protocol: TCP
      port: 80
Enter fullscreen mode Exit fullscreen mode

Then, will check the istio.enabled condition, and if it's set to the true - a Gateway and VirtualService resources will be created:

{{- if .Values.istio.enabled }}
--------
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: {{ .Values.appConfig.name }}-gateway
spec:     
  selector: 
    istio: ingressgateway
  servers:
  - port: 
      number: 80
      name: http
      protocol: HTTP
    hosts:  
    - {{ .Values.appConfig.url }}
--------         
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: {{ .Values.appConfig.name }}-virtualservice
spec:
  hosts:
  - {{ .Values.appConfig.url }}
  gateways:
  - {{ .Values.appConfig.name }}-gateway
  http:
  - match:
    - uri:
        prefix: /
    route:
    - destination:
        host: {{ .Values.appConfig.name }}-service
        port:
          number: 80
{{- end }}
Enter fullscreen mode Exit fullscreen mode

Ingress

For the Ingress in the ingress.enabled condition at first will check if need to create the Ingress at all, and if true - then will check which namespace to use, as if we will be using Istio, then this Ingress must be created in the istio-system namespace, and if it's a "common" Ingress - then in an application's namespace.

later, with the istio.enabled we will check where its traffic will be sent - to Istio or a common Service of the application:

{{- if .Values.ingress.enabled }}
-------- 
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: {{ .Values.appConfig.name }}-alb
  {{- if .Values.istio.enabled }}
  namespace: istio-system
  {{- end }}
  annotations:
    kubernetes.io/ingress.class: alb
    alb.ingress.kubernetes.io/scheme: internet-facing
    alb.ingress.kubernetes.io/certificate-arn: "arn:aws:acm:us-east-2:534***385:certificate/db886018-c173-43d0-b1be-b940de71f2a2"
    alb.ingress.kubernetes.io/listen-ports: '[{"HTTP": 80}, {"HTTPS":443}]'
    alb.ingress.kubernetes.io/actions.ssl-redirect: '{"Type": "redirect", "RedirectConfig": { "Protocol": "HTTPS", "Port": "443", "StatusCode": "HTTP_301"}}'
    external-dns.alpha.kubernetes.io/hostname: {{ .Values.appConfig.url }}
spec:   
  rules:
  - http:
      paths:
        - path: /*
          backend:
            serviceName: ssl-redirect
            servicePort: use-annotation
        - path: /*
          backend:
          {{- if .Values.istio.enabled }}
            serviceName: istio-ingressgateway
          {{ else }}
            serviceName: {{ .Values.appConfig.name }}-service
          {{- end }}
            servicePort: 80
{{- end }}
Enter fullscreen mode Exit fullscreen mode

Run Helm with the --debug and --dry-run - no Ingress must be printed, but a Gateway with VirtualService must be present:

$ helm upgrade — install backend-app-1 — namespace dev-backend-app-1-ns — create-namespace -f app-1/env/dev/values.yaml app-1/  -- debug --— dry-run
Enter fullscreen mode Exit fullscreen mode

If no errors here, then deploy it to the Dev cluster:

$ helm upgrade --install backend-app-1 --namespace dev-backend-app-1-ns --create-namespace -f app-1/env/dev/values.yaml app-1/
Release “backend-app-1” does not exist. Installing it now.
NAME: backend-app-1
LAST DEPLOYED: Mon Apr 12 18:03:08 2021
NAMESPACE: dev-backend-app-1-ns
STATUS: deployed
REVISION: 1
TEST SUITE: None
Enter fullscreen mode Exit fullscreen mode

Check an Ingress in the dev-backend-app-1-ns namespace:

$ kubectl -n dev-backend-app-1-ns get ingress
No resources found.
Enter fullscreen mode Exit fullscreen mode

Good, and we must see a Gateway and VirtualServce created:

$ kubectl -n dev-backend-app-1-ns get gateway
NAME AGE
backend-app-1-gateway 38s
kubectl -n dev-backend-app-1-ns get virtualservice
NAME GATEWAYS HOSTS AGE
backend-app-1-virtualservice [backend-app-1-gateway] [dev-app-1-test-common-ingress.example.com] 65s
Enter fullscreen mode Exit fullscreen mode

Good!

Now, deploy a “Production”:

$ helm upgrade --install backend-app-1 --namespace prod-backend-app-1-ns --create-namespace -f app-1/env/prod/values.yaml app-1/
Enter fullscreen mode Exit fullscreen mode

Check its Ingress in the prod-backend-app-1-ns namespace as we set do not use Istio here:

$ kubectl -n prod-backend-app-1-ns get ingress
NAME CLASS HOSTS ADDRESS PORTS AGE
backend-app-1-alb <none> * aadca942-***-49478225.us-east-2.elb.amazonaws.com 80 4m54s
Enter fullscreen mode Exit fullscreen mode

For the Dev environment, update the Ingress which we created at the very beginning — this is our shared Ingress. Add the external-dns.alpha.kubernetes.io/hostname here so ExternalDNS will create a record mapped to this LoadBalancer:

...
    # ExternalDNS settings: [https://rtfm.co.ua/en/kubernetes-update-aws-route53-dns-from-an-ingress/](https://rtfm.co.ua/en/kubernetes-update-aws-route53-dns-from-an-ingress/)
    external-dns.alpha.kubernetes.io/hostname: "dev-app-1-test-common-ingress.example.com"
...
Enter fullscreen mode Exit fullscreen mode

Apply:

$ kubectl apply -f common-ingress-gateway.yaml
ingress.extensions/backend-common-alb configured
Enter fullscreen mode Exit fullscreen mode

Check Istio Ingress Gateway routes — we must see only Dev routes for now:

$ istioctl proxy-config routes -n istio-system istio-ingressgateway-8459df68cb-bh76b — name http.80
NOTE: This output only contains routes loaded via RDS.
NAME DOMAINS MATCH VIRTUAL SERVICE
http.80 dev-app-1-test-common-ingress.example.com /* backend-app-1-virtualservice.dev-backend-app-1-ns
Enter fullscreen mode Exit fullscreen mode

Or in this way:

$ istioctl proxy-config routes -n istio-system istio-ingressgateway-8459df68cb-bh76b — name http.80 -o json | jq ‘.[].virtualHosts[].domains[0], .[].virtualHosts[].routes[].route.cluster’
“dev-app-1-test-common-ingress.example.com”
“outbound|80||backend-app-1-service.dev-backend-app-1-ns.svc.cluster.local”
Enter fullscreen mode Exit fullscreen mode

And try to access the application’s URLs.

Prod:

$ curl -I [https://prod-app-1-test-common-ingress.example.com](https://prod-app-1-test-common-ingress.example.com)
HTTP/2 200
date: Tue, 13 Apr 2021 12:47:15 GMT
content-type: text/html
server: nginx/1.13.8
Enter fullscreen mode Exit fullscreen mode

server: nginx/1.13.8 - the response received from the NGINX, and that means the request was sent via a LoadBalancer directly to the Service of the application to its Pod.

And Dev:

$ curl -I [https://dev-app-1-test-common-ingress.example.com](https://dev-app-1-test-common-ingress.example.com)
HTTP/2 200
date: Tue, 13 Apr 2021 12:47:18 GMT
content-type: text/html
server: istio-envoy
Enter fullscreen mode Exit fullscreen mode

server: istio-envoy - traffic went through the shared LoadBalancer, then to the Istio Ingress Gateway, then to a sidecar container with the Envoy proxy in the Pod with the application.

And now, let’s check the third available scheme — create a dedicated Ingress, but enable Istio for it.

In the app-1/env/prod/values.yaml change istio.enabled to the true, ingress.enabled we already have set to true:

appConfig:
  name: "backend-app-1"
  version: "v1"
  url: "prod-app-1-test-common-ingress.example.com"

istio:
  enabled: true

ingress:
  enabled: true
Enter fullscreen mode Exit fullscreen mode

Update the setup:

$ helm upgrade --install backend-app-1 --namespace prod-backend-app-1-ns --create-namespace -f app-1/env/prod/values.yaml app-1/
Enter fullscreen mode Exit fullscreen mode

Check Istio Ingress Gateway routes again:

$ istioctl proxy-config routes -n istio-system istio-ingressgateway-8459df68cb-bh76b --name http.80
NOTE: This output only contains routes loaded via RDS.
NAME DOMAINS MATCH VIRTUAL SERVICE
http.80 dev-app-1-test-common-ingress.example.com /* backend-app-1-virtualservice.dev-backend-app-1-ns
http.80 prod-app-1-test-common-ingress.example.com /* backend-app-1-virtualservice.prod-backend-app-1-ns
Enter fullscreen mode Exit fullscreen mode

Yup, we get a new route to the Production backend now.

Check the Ingress in the prod-backend-app-1-ns namespace:

$ kubectl -n prod-backend-app-1-ns get ingress backend-app-1-alb
Error from server (NotFound): ingresses.extensions “backend-app-1-alb” not found
Enter fullscreen mode Exit fullscreen mode

Good, and check the istio-system namespace:

$ kubectl -n istio-system get ingress backend-app-1-alb
NAME CLASS HOSTS ADDRESS PORTS AGE
backend-app-1-alb <none> * aadca942-***-1554475105.us-east-2.elb.amazonaws.com 80 8m52s
Enter fullscreen mode Exit fullscreen mode

Try with curl:

$ curl -I [https://prod-app-1-test-common-ingress.example.com](https://prod-app-1-test-common-ingress.example.com)
HTTP/2 200
date: Tue, 13 Apr 2021 13:14:34 GMT
content-type: text/html
server: istio-envoy
Enter fullscreen mode Exit fullscreen mode

server: istio-envoy - great! Our traffic now is going through the Istio.

Istio and ExternalDNS

And the last thing is to use ExternalDNS with Istio.

Currently, when using a shared Ingress and LoadBalancer, we can specify a Host/URL in the annotations of this Ingress, but this Ingress will not affected by Helm charts of the applications as it’s created from a dedicated manifest common-ingress-gateway.yaml.

So, to be able to create DNS records during application deployments we will need to update the annotations of the shared Ingress, and this leads to ad additional automation and complexity.

Instead, we can configure ExternalDNS in a way, when it will use not only Ingresses annotations, but Istio’s resources as well.

Let’s update its Deployment and add the --source=istio-gateway and/or --source=istio-virtualservice, see the documentation here>>>:

...
      containers:
      - args:
        - --log-level=info
        - --log-format=text
        - --events
        - --policy=upsert-only
        - --provider=aws
        - --registry=txt
        - --interval=2m
        - --source=service
        - --source=ingress
        - --source=istio-gateway
        - --source=istio-virtualservice
...
Enter fullscreen mode Exit fullscreen mode

From the common-ingress-gateway.yaml remove the line:

...
external-dns.alpha.kubernetes.io/hostname: "dev-app-1-test-common-ingress.example.com"
...
Enter fullscreen mode Exit fullscreen mode

Now, a hostname will be set in in a Gateway and/or a VirtualService, from the spec.servers.hosts for the Gateway or spec.hosts for VirtualService.

Also, check if ExternalDNS able to read Istio resources in its ClusterRole external-dns:

...
- apiGroups:
  - networking.istio.io
  resources:
  - virtualservices
  - gateways
  verbs:
  - get
  - list
  - watch
...
Enter fullscreen mode Exit fullscreen mode

ExternalDNS doesn’t create records for Istio Gateway and VirtualService: “No endpoints could be generated”

But here I’ve faced with an issue.

Enable --debug in the ExternalDNS Deployment and check its logs:

…
time=”2021–04–14T12:53:05Z” level=debug msg=”Adding event handler for Istio VirtualService”
time=”2021–04–14T12:53:05Z” level=debug msg=”Adding event handler for Istio Gateway”
time=”2021–04–14T12:53:05Z” level=debug msg=”Adding event handler for service”
time=”2021–04–14T12:53:05Z” level=debug msg=”Adding event handler for ingress”
…
level=debug msg=”No endpoints could be generated from ingress istio-system/backend-common-alb”
…
Enter fullscreen mode Exit fullscreen mode

Handlers for istio were created, so ExternalDNS is able to see Istio’s updates, but it can not create new records.

This happens due to the fact that Istio Ingress Gateway Service is created with the NodePort Service’s type to make it working with the shared LoadBalancer instead of the default LoadBalancer, and ExternalDNS can not parse a VirtualService to determine Kubernetes endpoints for an external Ingress, created from the common-ingress-gateway.yaml manifest.

We can “fix” it with a small hack: specify ALB’s URL directly in the VirtualService annotations.

Still, remember that in some cases a VirtualService will be created with a dedicated Ingress/ALB, and then we don’t need to add this annotation.

Therefore, add a new condition to the VirtualService manifest —  if not .Values.ingress.enabled:

apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: {{ .Values.appConfig.name }}-virtualservice
  {{- if not .Values.ingress.enabled }}
  annotations:
    external-dns.alpha.kubernetes.io/target: {{ .Values.istio.externalURL }}
  {{- end }}
...
Enter fullscreen mode Exit fullscreen mode

And set a value for the istio.externalURL in the app-1/env/dev/values.yaml file - it will be persistent enough and will be used for the Dev environment only:

...
istio:
  enabled: true
  externalURL: "aadca942-istiosystem-backe-3ee2-700661912.us-east-2.elb.amazonaws.com"
...
Enter fullscreen mode Exit fullscreen mode

Altogether, a Service, Gateway, and VirtualService will have such a manifest:

-------- 
apiVersion: v1
kind: Service
metadata:
  name: {{ .Values.appConfig.name }}-service
spec:
  {{- if .Values.ingress.enabled }}
  type: NodePort
  {{- end }}
  selector:
    app: {{ .Values.appConfig.name }}
  ports:
    - name: http
      protocol: TCP
      port: 80

{{- if .Values.istio.enabled }}
-------- 
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: {{ .Values.appConfig.name }}-gateway
spec:
  selector:
    istio: ingressgateway
  servers:  
  - port:
      number: 80
      name: http 
      protocol: HTTP
    hosts:
    - "*"
--------       
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: {{ .Values.appConfig.name }}-virtualservice
  {{- if not .Values.ingress.enabled }}
  annotations:
    external-dns.alpha.kubernetes.io/target: {{ .Values.istio.externalURL }}
  {{- end }}
spec:
  hosts:
  - {{ .Values.appConfig.url | quote }}
  gateways:
  - {{ .Values.appConfig.name }}-gateway
  http:
  - match:
    - uri:
        prefix: /
    route:
    - destination:
        host: {{ .Values.appConfig.name }}-service
        port:
          number: 80
{{- end }}
Enter fullscreen mode Exit fullscreen mode

Deploy it and check ExternalDNS logs:

…
time=”2021–04–14T13:05:00Z” level=info msg=”Desired change: CREATE dev-app-1-test-common-ingress.example.com A [Id: /hostedzone/Z30KLN6M3D0LB6]”
time=”2021–04–14T13:05:00Z” level=info msg=”Desired change: CREATE dev-app-1-test-common-ingress.example.com TXT [Id: /hostedzone/Z30KLN6M3D0LB6]”
…
Enter fullscreen mode Exit fullscreen mode

Try again with curl:

$ curl -I [https://dev-app-1-test-common-ingress.example.com](https://dev-app-1-test-common-ingress.example.com)
HTTP/2 200
date: Wed, 14 Apr 2021 13:21:11 GMT
content-type: text/html
server: istio-envoy
Enter fullscreen mode Exit fullscreen mode

All done.

Originally published at RTFM: Linux, DevOps, and system administration.


Top comments (0)