I'm trying to configure a canary rollout for a demo, but I'm having trouble getting the traffic splitting to work with linkerd. The funny part is I was able to get this working with istio and i find istio to be much more complicated then linkerd.
I have a basic go-lang service define like this:
apiVersion: argoproj.io/v1alpha1
kind: Rollout
metadata:
name: fish
spec:
[...]
strategy:
canary:
canaryService: canary-svc
stableService: stable-svc
trafficRouting:
smi: {}
steps:
- setWeight: 5
- pause: {}
- setWeight: 20
- pause: {}
- setWeight: 50
- pause: {}
- setWeight: 80
- pause: {}
---
apiVersion: v1
kind: Service
metadata:
name: canary-svc
spec:
selector:
app: fish
ports:
- name: http
port: 8080
---
apiVersion: v1
kind: Service
metadata:
name: stable-svc
spec:
selector:
app: fish
ports:
- name: http
port: 8080
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: fish
annotations:
kubernetes.io/ingress.class: 'nginx'
cert-manager.io/cluster-issuer: letsencrypt-production
cert-manager.io/acme-challenge-type: dns01
external-dns.alpha.kubernetes.io/hostname: fish.local
nginx.ingress.kubernetes.io/enable-cors: "true"
nginx.ingress.kubernetes.io/cors-allow-methods: "PUT, GET, POST, OPTIONS"
nginx.ingress.kubernetes.io/cors-allow-origin: "*"
nginx.ingress.kubernetes.io/configuration-snippet: |
proxy_set_header l5d-dst-override $service_name.$namespace.svc.cluster.local:$service_port;
spec:
rules:
- host: fish.local
http:
paths:
- path: /
pathType: Prefix
backend:
service:
name: stable-svc
port:
number: 8080
When I do the deploy (sync) via ArgoCD I can see the traffic split is 50/50:
- apiVersion: split.smi-spec.io/v1alpha2
kind: TrafficSplit
metadata:
[...]
name: fish
namespace: default
spec:
backends:
- service: canary-svc
weight: "50"
- service: stable-svc
weight: "50"
service: stable-svc
However doing a curl command in a while loop i only get back the stable-svc. The only time i see a change is after I have completely moved the service to 100%.
I tried to follow this: https://argoproj.github.io/argo-rollouts/getting-started/smi/
Any help would be greatly appreciated.
Thanks
After reading this: https://linkerd.io/2.10/tasks/using-ingress/ I discovered you need to modify your ingress controller with a special annotation:
$ kubectl get deployment <ingress-controller> -n <ingress-namespace> -o yaml | linkerd inject --ingress - | kubectl apply -f -
TLDR; if you want Linkerd functionality like Service Profiles, Traffic Splits, etc, there is additional configuration required to make the Ingress controller’s Linkerd proxy run in ingress mode.
So there's a bit more context in this issue but the TL;DR is ingresses tend to target individual pods instead of the service address. Putting Linkerd's proxy in ingress mode tells it to override that behaviour. NGINX does already have a setting that will let it hit services instead of endpoints directly, you can see that in their docs here.
I'm getting below error when trying to reach an internal neo4j endpoint from another cluster
neobolt.exceptions.ServiceUnavailable: Timed out trying to establish connection to ('xx.xxx.xx.xx', 7687)
When accessing this endpoint through the browser, it shows
not a WebSocket handshake request: missing upgrade
I work on GCP. This is what I've got:
Cluster A with Composer running Airflow
Cluster B with K8s where my application is deployed
I know for sure both clusters can communicate
Cluster B has a neo4j ingress defined as follows:
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
annotations:
field.cattle.io/publicEndpoints: '[{"addresses":["xx.xxx.xx.xx"],"port":443,"protocol":"HTTPS","serviceName":"dev:neo4j","ingressName":"dev:neo4j-dev-ing","hostname":"neo4j-dev.host_name","allNodes":false}]'
generation: 6
name: neo4j-dev-ing
spec:
rules:
- host: neo4j-dev.host_name
http:
paths:
- backend:
serviceName: neo4j
servicePort: neo4j-dev-bolt
tls:
- hosts:
- neo4j-dev.host_name
status:
loadBalancer:
ingress:
- ip: xx.xxx.xx.xx
My neo4j service looks as follows:
apiVersion: v1
kind: Service
metadata:
name: neo4j
spec:
type: ClusterIP
selector:
app: neo4j
component: neo4j
ports:
- port: 7473
name: neo4j-dev-https
targetPort: 7473
- port: 7474
name: neo4j-dev-http
targetPort: 7474
- port: 7687
name: neo4j-dev-bolt
targetPort: 7687
- port: 1337
name: neo4j-dev-shell
targetPort: 1337
I've seen a few related questions but nothing concrete and a lot of contradictory information.
Any ideas if this can even work at all? Can someone with some insight into networks explain to me why this isn't working or put forward the concepts I need to research in order to get what's going on? Got stuck
It’s a known issue with accessing neo4j outside of kubernetes and there are no straightforward workarounds
A complex workaround using multiple static IPs has been described here
https://neo4j.com/labs/neo4j-helm/1.0.0/externalexposure/
I have Pod and Service ymal files in my system. I want to run these two using kubectl create -f <file> and connect from outside browser to test connectivity.Here what I have followed.
My Pod :
apiVersion: v1
kind: Pod
metadata:
name: client-nginx
labels:
component: web
spec:
containers:
- name: client
image: nginx
ports:
- containerPort: 3000
My Services file :
apiVersion: v1
kind: Service
metadata:
name: client-nginx-port
spec:
type: NodePort
ports:
- port: 3050
targetPort: 3000
nodePort: 31616
selector:
component: web
I used kubectl create -f my_pod.yaml and then kubectl get pods shows my pod client-nginx
And then kubectl create -f my_service.yaml, No errors here and then shows all the services.
When I try to curl to service, it gives
curl: (7) Failed to connect to 192.168.0.10 port 31616: Connection refused.
kubectl get deployments doesnt show my pod. Do I have to deploy it? I am a bit confused. If I use instructions given here, I can deploynginxsuccessfully and access from outside browsers.
I used instructions given here to test this.
Try with this service:
apiVersion: v1
kind: Service
metadata:
name: client-nginx-port
spec:
type: NodePort
ports:
- port: 3050
targetPort: 80
nodePort: 31616
selector:
component: web
You missed selector name to be given to pod yaml which will be picked by service where you have mentioned the selector as component
Use this in pod yaml
apiVersion: v1
kind: Pod
metadata:
name: client-nginx
labels:
component: web
spec:
selector:
component: nginx
containers:
- name: client
image: nginx
ports:
- containerPort: 3000
Useful links:
https://kubernetes.io/docs/concepts/services-networking/service/
https://kubernetes.io/docs/concepts/services-networking/connect-applications-service/
I've been banging my head against this wall on and off for a while. There is a ton of information on Kubernetes on the web, but it's all assuming so much knowledge that n00bs like me don't really have much to go on.
So, can anyone share a simple example of the following (as a yaml file)? All I want is
two pods
let's say one pod has a backend (I don't know - node.js), and one has a frontend (say React).
A way to network between them.
And then an example of calling an api call from the back to the front.
I start looking into this sort of thing, and all of a sudden I hit this page - https://kubernetes.io/docs/concepts/cluster-administration/networking/#how-to-achieve-this. This is super unhelpful. I don't want or need advanced network policies, nor do I have the time to go through several different service layers that are mapped on top of kubernetes. I just want to figure out a trivial example of a network request.
Hopefully if this example exists on stackoverflow it will serve other people as well.
Any help would be appreciated. Thanks.
EDIT; it looks like the easiest example may be using the Ingress controller.
EDIT EDIT;
I'm working to try and get a minimal example deployed - I'll walk through some steps here and point out my issues.
So below is my yaml file:
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: frontend
labels:
app: frontend
spec:
replicas: 3
selector:
matchLabels:
app: frontend
template:
metadata:
labels:
app: frontend
spec:
containers:
- name: nginx
image: patientplatypus/frontend_example
ports:
- containerPort: 3000
---
apiVersion: v1
kind: Service
metadata:
name: frontend
spec:
type: LoadBalancer
selector:
app: frontend
ports:
- protocol: TCP
port: 80
targetPort: 3000
---
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: backend
labels:
app: backend
spec:
replicas: 3
selector:
matchLabels:
app: backend
template:
metadata:
labels:
app: backend
spec:
containers:
- name: nginx
image: patientplatypus/backend_example
ports:
- containerPort: 5000
---
apiVersion: v1
kind: Service
metadata:
name: backend
spec:
type: LoadBalancer
selector:
app: backend
ports:
- protocol: TCP
port: 80
targetPort: 5000
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
name: frontend
spec:
rules:
- host: www.kubeplaytime.example
http:
paths:
- path: /
backend:
serviceName: frontend
servicePort: 80
- path: /api
backend:
serviceName: backend
servicePort: 80
What I believe this is doing is
Deploying a frontend and backend app - I deployed patientplatypus/frontend_example and patientplatypus/backend_example to dockerhub and then pull the images down. One open question I have is, what if I don't want to pull the images from docker hub and rather would just like to load from my localhost, is that possible? In this case I would push my code to the production server, build the docker images on the server and then upload to kubernetes. The benefit is that I don't have to rely on dockerhub if I want my images to be private.
It is creating two service endpoints that route outside traffic from a web browser to each of the deployments. These services are of type loadBalancer because they are balancing the traffic among the (in this case 3) replicasets that I have in the deployments.
Finally, I have an ingress controller which is supposed to allow my services to route to each other through www.kubeplaytime.example and www.kubeplaytime.example/api. However this is not working.
What happens when I run this?
patientplatypus:~/Documents/kubePlay:09:17:50$kubectl create -f kube-deploy.yaml
deployment.apps "frontend" created
service "frontend" created
deployment.apps "backend" created
service "backend" created
ingress.extensions "frontend" created
So first, it appears to create all the parts that I need fine with no errors.
patientplatypus:~/Documents/kubePlay:09:22:30$kubectl get --watch services
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
backend LoadBalancer 10.0.18.174 <pending> 80:31649/TCP 1m
frontend LoadBalancer 10.0.100.65 <pending> 80:32635/TCP 1m
kubernetes ClusterIP 10.0.0.1 <none> 443/TCP 10d
frontend LoadBalancer 10.0.100.65 138.91.126.178 80:32635/TCP 2m
backend LoadBalancer 10.0.18.174 138.91.121.182 80:31649/TCP 2m
Second, if I watch the services, I eventually get IP addresses that I can use to navigate in my browser to these sites. Each of the above IP addresses works in routing me to the frontend and backend respectively.
HOWEVER
I reach an issue when I try and use the ingress controller - it seemingly deployed, but I don't know how to get there.
patientplatypus:~/Documents/kubePlay:09:24:44$kubectl get ingresses
NAME HOSTS ADDRESS PORTS AGE
frontend www.kubeplaytime.example 80 16m
So I have no address I can use, and www.kubeplaytime.example does not appear to work.
What it appears that I have to do to route to the ingress extension I just created is to use a service and deployment on it in order to get an IP address, but this starts to look incredibly complicated very quickly.
For example, take a look at this medium article: https://medium.com/#cashisclay/kubernetes-ingress-82aa960f658e.
It would appear that the necessary code to add for just the service routing to the Ingress (ie what he calls the Ingress Controller) appears to be this:
---
kind: Service
apiVersion: v1
metadata:
name: ingress-nginx
spec:
type: LoadBalancer
selector:
app: ingress-nginx
ports:
- name: http
port: 80
targetPort: http
- name: https
port: 443
targetPort: https
---
kind: Deployment
apiVersion: extensions/v1beta1
metadata:
name: ingress-nginx
spec:
replicas: 1
template:
metadata:
labels:
app: ingress-nginx
spec:
terminationGracePeriodSeconds: 60
containers:
- image: gcr.io/google_containers/nginx-ingress-controller:0.8.3
name: ingress-nginx
imagePullPolicy: Always
ports:
- name: http
containerPort: 80
protocol: TCP
- name: https
containerPort: 443
protocol: TCP
livenessProbe:
httpGet:
path: /healthz
port: 10254
scheme: HTTP
initialDelaySeconds: 30
timeoutSeconds: 5
env:
- name: POD_NAME
valueFrom:
fieldRef:
fieldPath: metadata.name
- name: POD_NAMESPACE
valueFrom:
fieldRef:
fieldPath: metadata.namespace
args:
- /nginx-ingress-controller
- --default-backend-service=$(POD_NAMESPACE)/nginx-default-backend
---
kind: Service
apiVersion: v1
metadata:
name: nginx-default-backend
spec:
ports:
- port: 80
targetPort: http
selector:
app: nginx-default-backend
---
kind: Deployment
apiVersion: extensions/v1beta1
metadata:
name: nginx-default-backend
spec:
replicas: 1
template:
metadata:
labels:
app: nginx-default-backend
spec:
terminationGracePeriodSeconds: 60
containers:
- name: default-http-backend
image: gcr.io/google_containers/defaultbackend:1.0
livenessProbe:
httpGet:
path: /healthz
port: 8080
scheme: HTTP
initialDelaySeconds: 30
timeoutSeconds: 5
resources:
limits:
cpu: 10m
memory: 20Mi
requests:
cpu: 10m
memory: 20Mi
ports:
- name: http
containerPort: 8080
protocol: TCP
This would seemingly need to be appended to my other yaml code above in order to get a service entry point for my ingress routing, and it does appear to give an ip:
patientplatypus:~/Documents/kubePlay:09:54:12$kubectl get --watch services
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
backend LoadBalancer 10.0.31.209 <pending> 80:32428/TCP 4m
frontend LoadBalancer 10.0.222.47 <pending> 80:32482/TCP 4m
ingress-nginx LoadBalancer 10.0.28.157 <pending> 80:30573/TCP,443:30802/TCP 4m
kubernetes ClusterIP 10.0.0.1 <none> 443/TCP 10d
nginx-default-backend ClusterIP 10.0.71.121 <none> 80/TCP 4m
frontend LoadBalancer 10.0.222.47 40.121.7.66 80:32482/TCP 5m
ingress-nginx LoadBalancer 10.0.28.157 40.121.6.179 80:30573/TCP,443:30802/TCP 6m
backend LoadBalancer 10.0.31.209 40.117.248.73 80:32428/TCP 7m
So ingress-nginx appears to be the site I want to get to. Navigating to 40.121.6.179 returns a default 404 message (default backend - 404) - it does not go to frontend as / aught to route. /api returns the same. Navigating to my host namespace www.kubeplaytime.example returns a 404 from the browser - no error handling.
QUESTIONS
Is the Ingress Controller strictly necessary, and if so is there a less complicated version of this?
I feel I am close, what am I doing wrong?
FULL YAML
Available here: https://gist.github.com/patientplatypus/fa07648339ee6538616cb69282a84938
Thanks for the help!
EDIT EDIT EDIT
I've attempted to use HELM. On the surface it appears to be a simple interface, and so I tried spinning it up:
patientplatypus:~/Documents/kubePlay:12:13:00$helm install stable/nginx-ingress
NAME: erstwhile-beetle
LAST DEPLOYED: Sun May 6 12:13:30 2018
NAMESPACE: default
STATUS: DEPLOYED
RESOURCES:
==> v1/ConfigMap
NAME DATA AGE
erstwhile-beetle-nginx-ingress-controller 1 1s
==> v1/Service
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
erstwhile-beetle-nginx-ingress-controller LoadBalancer 10.0.216.38 <pending> 80:31494/TCP,443:32118/TCP 1s
erstwhile-beetle-nginx-ingress-default-backend ClusterIP 10.0.55.224 <none> 80/TCP 1s
==> v1beta1/Deployment
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
erstwhile-beetle-nginx-ingress-controller 1 1 1 0 1s
erstwhile-beetle-nginx-ingress-default-backend 1 1 1 0 1s
==> v1beta1/PodDisruptionBudget
NAME MIN AVAILABLE MAX UNAVAILABLE ALLOWED DISRUPTIONS AGE
erstwhile-beetle-nginx-ingress-controller 1 N/A 0 1s
erstwhile-beetle-nginx-ingress-default-backend 1 N/A 0 1s
==> v1/Pod(related)
NAME READY STATUS RESTARTS AGE
erstwhile-beetle-nginx-ingress-controller-7df9b78b64-24hwz 0/1 ContainerCreating 0 1s
erstwhile-beetle-nginx-ingress-default-backend-849b8df477-gzv8w 0/1 ContainerCreating 0 1s
NOTES:
The nginx-ingress controller has been installed.
It may take a few minutes for the LoadBalancer IP to be available.
You can watch the status by running 'kubectl --namespace default get services -o wide -w erstwhile-beetle-nginx-ingress-controller'
An example Ingress that makes use of the controller:
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
annotations:
kubernetes.io/ingress.class: nginx
name: example
namespace: foo
spec:
rules:
- host: www.example.com
http:
paths:
- backend:
serviceName: exampleService
servicePort: 80
path: /
# This section is only required if TLS is to be enabled for the Ingress
tls:
- hosts:
- www.example.com
secretName: example-tls
If TLS is enabled for the Ingress, a Secret containing the certificate and key must also be provided:
apiVersion: v1
kind: Secret
metadata:
name: example-tls
namespace: foo
data:
tls.crt: <base64 encoded cert>
tls.key: <base64 encoded key>
type: kubernetes.io/tls
Seemingly this is really nice - it spins everything up and gives an example of how to add an ingress. Since I spun up helm in a blank kubectl I used the following yaml file to add in what I thought would be required.
The file:
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: frontend
labels:
app: frontend
spec:
replicas: 3
selector:
matchLabels:
app: frontend
template:
metadata:
labels:
app: frontend
spec:
containers:
- name: nginx
image: patientplatypus/frontend_example
ports:
- containerPort: 3000
---
apiVersion: v1
kind: Service
metadata:
name: frontend
spec:
type: LoadBalancer
selector:
app: frontend
ports:
- protocol: TCP
port: 80
targetPort: 3000
---
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: backend
labels:
app: backend
spec:
replicas: 3
selector:
matchLabels:
app: backend
template:
metadata:
labels:
app: backend
spec:
containers:
- name: nginx
image: patientplatypus/backend_example
ports:
- containerPort: 5000
---
apiVersion: v1
kind: Service
metadata:
name: backend
spec:
type: LoadBalancer
selector:
app: backend
ports:
- protocol: TCP
port: 80
targetPort: 5000
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
annotations:
kubernetes.io/ingress.class: nginx
spec:
rules:
- host: www.example.com
http:
paths:
- path: /api
backend:
serviceName: backend
servicePort: 80
- path: /
frontend:
serviceName: frontend
servicePort: 80
Deploying this to the cluster however runs into this error:
patientplatypus:~/Documents/kubePlay:11:44:20$kubectl create -f kube-deploy.yaml
deployment.apps "frontend" created
service "frontend" created
deployment.apps "backend" created
service "backend" created
error: error validating "kube-deploy.yaml": error validating data: [ValidationError(Ingress.spec.rules[0].http.paths[1]): unknown field "frontend" in io.k8s.api.extensions.v1beta1.HTTPIngressPath, ValidationError(Ingress.spec.rules[0].http.paths[1]): missing required field "backend" in io.k8s.api.extensions.v1beta1.HTTPIngressPath]; if you choose to ignore these errors, turn validation off with --validate=false
So, the question then becomes, well crap how do I debug this?
If you spit out the code that helm produces, it's basically non-readable by a person - there's no way to go in there and figure out what's going on.
Check it out: https://gist.github.com/patientplatypus/0e281bf61307f02e16e0091397a1d863 - over a 1000 lines!
If anyone has a better way to debug a helm deploy add it to the list of open questions.
EDIT EDIT EDIT EDIT
To simplify in the extreme I attempt to make a call from one pod to another only using namespace.
So here is my React code where I make the http request:
axios.get('http://backend/test')
.then(response=>{
console.log('return from backend and response: ', response);
})
.catch(error=>{
console.log('return from backend and error: ', error);
})
I've also attempted to use http://backend.exampledeploy.svc.cluster.local/test without luck.
Here is my node code handling the get:
router.get('/test', function(req, res, next) {
res.json({"test":"test"})
});
Here is my yaml file that I uploading to the kubectl cluster:
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: frontend
namespace: exampledeploy
labels:
app: frontend
spec:
replicas: 3
selector:
matchLabels:
app: frontend
template:
metadata:
labels:
app: frontend
spec:
containers:
- name: nginx
image: patientplatypus/frontend_example
ports:
- containerPort: 3000
---
apiVersion: v1
kind: Service
metadata:
name: frontend
namespace: exampledeploy
spec:
type: LoadBalancer
selector:
app: frontend
ports:
- protocol: TCP
port: 80
targetPort: 3000
---
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: backend
namespace: exampledeploy
labels:
app: backend
spec:
replicas: 3
selector:
matchLabels:
app: backend
template:
metadata:
labels:
app: backend
spec:
containers:
- name: nginx
image: patientplatypus/backend_example
ports:
- containerPort: 5000
---
apiVersion: v1
kind: Service
metadata:
name: backend
namespace: exampledeploy
spec:
type: LoadBalancer
selector:
app: backend
ports:
- protocol: TCP
port: 80
targetPort: 5000
The uploading to the cluster appears to work as I can see in my terminal:
patientplatypus:~/Documents/kubePlay:14:33:20$kubectl get all --namespace=exampledeploy
NAME READY STATUS RESTARTS AGE
pod/backend-584c5c59bc-5wkb4 1/1 Running 0 15m
pod/backend-584c5c59bc-jsr4m 1/1 Running 0 15m
pod/backend-584c5c59bc-txgw5 1/1 Running 0 15m
pod/frontend-647c99cdcf-2mmvn 1/1 Running 0 15m
pod/frontend-647c99cdcf-79sq5 1/1 Running 0 15m
pod/frontend-647c99cdcf-r5bvg 1/1 Running 0 15m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/backend LoadBalancer 10.0.112.160 168.62.175.155 80:31498/TCP 15m
service/frontend LoadBalancer 10.0.246.212 168.62.37.100 80:31139/TCP 15m
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deployment.extensions/backend 3 3 3 3 15m
deployment.extensions/frontend 3 3 3 3 15m
NAME DESIRED CURRENT READY AGE
replicaset.extensions/backend-584c5c59bc 3 3 3 15m
replicaset.extensions/frontend-647c99cdcf 3 3 3 15m
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deployment.apps/backend 3 3 3 3 15m
deployment.apps/frontend 3 3 3 3 15m
NAME DESIRED CURRENT READY AGE
replicaset.apps/backend-584c5c59bc 3 3 3 15m
replicaset.apps/frontend-647c99cdcf 3 3 3 15m
However, when I attempt to make the request I get the following error:
return from backend and error:
Error: Network Error
Stack trace:
createError#http://168.62.37.100/static/js/bundle.js:1555:15
handleError#http://168.62.37.100/static/js/bundle.js:1091:14
App.js:14
Since the axios call is being made from the browser, I'm wondering if it is simply not possible to use this method to call the backend, even though the backend and the frontend are in different pods. I'm a little lost, as I thought this was the simplest possible way to network pods together.
EDIT X5
I've determined that it is possible to curl the backend from the command line by exec'ing into the pod like this:
patientplatypus:~/Documents/kubePlay:15:25:25$kubectl exec -ti frontend-647c99cdcf-5mfz4 --namespace=exampledeploy -- curl -v http://backend/test
* Hostname was NOT found in DNS cache
* Trying 10.0.249.147...
* Connected to backend (10.0.249.147) port 80 (#0)
> GET /test HTTP/1.1
> User-Agent: curl/7.38.0
> Host: backend
> Accept: */*
>
< HTTP/1.1 200 OK
< X-Powered-By: Express
< Content-Type: application/json; charset=utf-8
< Content-Length: 15
< ETag: W/"f-SzkCEKs7NV6rxiz4/VbpzPnLKEM"
< Date: Sun, 06 May 2018 20:25:49 GMT
< Connection: keep-alive
<
* Connection #0 to host backend left intact
{"test":"test"}
What this means is, without a doubt, because the front end code is being executed in the browser it needs Ingress to gain entry into the pod, as http requests from the front end are what's breaking with simple pod networking. I was unsure of this, but it means Ingress is necessary.
First of all, let's clarify some apparent misconceptions. You mentioned your front-end being a React application, that will presumably run in the users browser. For this to work, your actual problem is not your back-end and front-end pods communicating with each other, but the browser needs to be able to connect to both these pods (to the front-end pod in order to load the React application, and to the back-end pod for the React app to make API calls).
To visualize:
+---------+
+---| Browser |---+
| +---------+ |
V V
+-----------+ +----------+ +-----------+ +----------+
| Front-end |---->| Back-end | | Front-end | | Back-end |
+-----------+ +----------+ +-----------+ +----------+
(what you asked for) (what you need)
As already stated, the easiest solution for this would be to use an Ingress controller. I won't go into detail on how to set up an Ingress controller here; in some cloud environments (like GKE) you will be able to use an Ingress controller provided to you by the cloud provider. Otherwise, you can set up the NGINX Ingress controller. Have a look at the NGINX Ingress controllers deployment guide for more information.
Define services
Start by defining Service resources for both your front-end and back-end application (these would also allow your Pods to communicate with each other). A service definition might look like this:
apiVersion: v1
kind: Service
metadata:
name: backend
spec:
selector:
app: backend
ports:
- protocol: TCP
port: 80
targetPort: 8080
Make sure that your Pods have labels that can be selected by the Service resource (in this example, I'm using app=backend and app=frontend as labels).
If you want to establish Pod-to-Pod communication, you're done now. In each Pod, you can now use backend.<namespace>.svc.cluster.local (or backend as shorthand) and frontend as host names to connect to that Pod.
Define Ingresses
Next up, you can define the Ingress resources; since both services will need connectivity from outside the cluster (the users browser), you will need Ingress definitions for both services.
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
name: frontend
spec:
rules:
- host: www.your-application.example
http:
paths:
- path: /
backend:
serviceName: frontend
servicePort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
name: backend
spec:
rules:
- host: api.your-application.example
http:
paths:
- path: /
backend:
serviceName: backend
servicePort: 80
Alternatively, you could also aggregate frontend and backend with a single Ingress resource (no "right" answer here, just a matter of preference):
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
name: frontend
spec:
rules:
- host: www.your-application.example
http:
paths:
- path: /
backend:
serviceName: frontend
servicePort: 80
- path: /api
backend:
serviceName: backend
servicePort: 80
After that, make sure that both www.your-application.example and api.your-application.example point to your Ingress controller's external IP address, and you should be done.
As it turns out I was over-complicating things. Here is the Kubernetes file that works to do what I want. You can do this using two deployments (front end, and backend) and one service entrypoint. As far as I can tell, a service can load balance to many (not just 2) different deployments, meaning for practical development this should be a good start to micro service development. One of the benefits of an ingress method is allowing the use of path names rather than port numbers, but given the difficulty it doesn't seem practical in development.
Here is the yaml file:
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: frontend
labels:
app: exampleapp
spec:
replicas: 3
selector:
matchLabels:
app: exampleapp
template:
metadata:
labels:
app: exampleapp
spec:
containers:
- name: nginx
image: patientplatypus/kubeplayfrontend
ports:
- containerPort: 3000
---
apiVersion: apps/v1beta1
kind: Deployment
metadata:
name: backend
labels:
app: exampleapp
spec:
replicas: 3
selector:
matchLabels:
app: exampleapp
template:
metadata:
labels:
app: exampleapp
spec:
containers:
- name: nginx
image: patientplatypus/kubeplaybackend
ports:
- containerPort: 5000
---
apiVersion: v1
kind: Service
metadata:
name: entrypt
spec:
type: LoadBalancer
ports:
- name: backend
port: 8080
targetPort: 5000
- name: frontend
port: 81
targetPort: 3000
selector:
app: exampleapp
Here are the bash commands I use to get it to spin up (you may have to add a login command - docker login - to push to dockerhub):
#!/bin/bash
# stop all containers
echo stopping all containers
docker stop $(docker ps -aq)
# remove all containers
echo removing all containers
docker rm $(docker ps -aq)
# remove all images
echo removing all images
docker rmi $(docker images -q)
echo building backend
cd ./backend
docker build -t patientplatypus/kubeplaybackend .
echo push backend to dockerhub
docker push patientplatypus/kubeplaybackend:latest
echo building frontend
cd ../frontend
docker build -t patientplatypus/kubeplayfrontend .
echo push backend to dockerhub
docker push patientplatypus/kubeplayfrontend:latest
echo now working on kubectl
cd ..
echo deleting previous variables
kubectl delete pods,deployments,services entrypt backend frontend
echo creating deployment
kubectl create -f kube-deploy.yaml
echo watching services spin up
kubectl get services --watch
The actual code is just a frontend react app making an axios http call to a backend node route on componentDidMount of the starting App page.
You can also see a working example here: https://github.com/patientplatypus/KubernetesMultiPodCommunication
Thanks again everyone for your help.
To use ingress controller you need to have valid domain (DNS server configured to point your ingress controller ip). This is not due to any kubernetes "magic" but due to the way how vhosts work (here is an example for nginx - very often used as ingress server, but any other ingress implementation will work the same way under the hood).
If you can't configure your domain the easiest way for dev purpose would be creating kubernetes service. There is a nice short cut for doing it using kubectl expose
kubectl expose pod frontend-pod --port=444 --name=frontend
kubectl expose pod backend-pod --port=888 --name=backend
I am deploying java application at google kubernetes engine. Application correctly starts but fails when trying to request data. Exception is "DatastoreException, Missing or insufficient permissions". I created service account with "Owner" role and provided service account key to kubernetes. Here is how i apply kubernetes deployment:
# delete old secret
kubectl delete secret google-key --ignore-not-found
# file with key
kubectl create secret generic google-key --from-file=key.json
kubectl apply -f prod-kubernetes.yml
Here is deployment config:
apiVersion: v1
kind: Service
metadata:
annotations:
service.alpha.kubernetes.io/tolerate-unready-endpoints: "true"
name: user
labels:
app: user
spec:
type: NodePort
ports:
- port: 8000
name: user
targetPort: 8000
nodePort: 32756
selector:
app: user
---
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
name: userdeployment
spec:
replicas: 1
template:
metadata:
labels:
app: user
spec:
volumes:
- name: google-cloud-key
secret:
secretName: google-key
containers:
- name: usercontainer
image: gcr.io/proj/user:v1
imagePullPolicy: Always
volumeMounts:
- name: google-cloud-key
mountPath: /var/secrets/google
env:
- name: GOOGLE_APPLICATION_CREDENTIALS
value: /var/secrets/google/key.json
ports:
- containerPort: 8000
I wonder why it is not working? I have used this config in previous deployment and had success.
UPD: I made sure that /var/secrets/google/key.json exist at pod. I print Files.exists(System.getEnv("GOOGLE_APPLICATION_CREDENTIALS")) to log. I also print content of this file - it seems not corrupted.
Solved, reason was incorrect evn name GOOGLE_CLOUD_PROJECT