gke/README.md

Deploy SmaRP on GKE with HTTP(S) load balancing

This document shows how to deploy the SmaRP Shiny app to GKE and set up an HTTPS load balancer exposing it through a URL, usable for embedding in website gallery.

Initialization

Here we assume that a deployment smarp already exists for SmaRP, running in the smarp cluster.

Check if the smarp cluster is the current context

kubectl config current-context
## gke_mirai-sbb_europe-west1-b_smarp

You can set it otherwise by running

gcloud container clusters get-credentials smarp
kubectl config current-context
## Fetching cluster endpoint and auth data.
## kubeconfig entry generated for smarp.
## gke_mirai-sbb_europe-west1-b_smarp

Query the available pods to check that smarp is deployed and running

kubectl get pods -o wide
## NAME                    READY   STATUS    RESTARTS   AGE    IP         NODE                                   NOMINATED NODE   READINESS GATES
## smarp-57b6df7b6-hfmss   1/1     Running   24         7d1h   10.4.0.9   gke-smarp-default-pool-90e87a82-gdg4   <none>           <none>

Setup and HTTP(s) load balancer

References

Introduction

HTTP(S) load balancing in GKE is based on three major components

Backend configuration

We need to use a backend configuration to specify a sensible maximum connection timeout, higher than the default 30s. Note that this is a typical situation for WebSockets. A sensible value is 10800s (3h), since it is also Kubernetes’ default max session sticky time for the “ClientIP” sessionAffinity.

See Configuring a backend service through Ingress for more details. Note that the BackendConfig custom resource is in a Beta release at the time of writing.

The corresponding BackendConfig YAML manifest, including session affinity, is as follows

## # smarp-backendconfig.yaml
## apiVersion: cloud.google.com/v1beta1
## kind: BackendConfig
## metadata:
##   name: smarp-backendconfig
## spec:
##   timeoutSec: 10800 # 3h
##   sessionAffinity:
##     affinityType: "CLIENT_IP"

The BackendConfig is then created from the manifest as

kubectl apply -f smarp-backendconfig.yaml
kubectl get backendconfig smarp-backendconfig
## NAME                  AGE
## smarp-backendconfig   9d

Expose your Deployment as a Service

The NodePort service manifest using the backend configuration created above is as follows

## # smarp-backend.yaml
## apiVersion: v1
## kind: Service
## metadata:
##   labels:
##     run: smarp
##   name: smarp-backend
##   annotations:
##     beta.cloud.google.com/backend-config: '{"ports": {"80":"smarp-backendconfig"}}'
## spec:
##   type: NodePort
##   selector:
##     run: smarp
##   ports:
##   - port: 80
##     targetPort: 80

TCP port 80 of the created Service is associated with a BackendConfig named smarp-backendconfig via annotation beta.cloud.google.com/backend-config. A request sent to port 80 of the service is forwarded to one of the member Pods on targetPort 80. Note that each member Pod must have a container listening on the specified targetPort. Note that there is no need to set the cloud.google.com/app-protocols annotation to handle requests to port 443 for the HTTPS protocol, since our app is not capable of receiving HTTPS requests.

An easy way to get the labels and selector right is to look at the YAML generated for a standard NodePort service using expose

kubectl expose deployment smarp --target-port=80 --port=80 --type=NodePort --dry-run -o=yaml

The corresponding service is then created from the manifest as

kubectl apply -f smarp-backend.yaml
kubectl get service smarp-backend
## NAME            TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)        AGE
## smarp-backend   NodePort   10.7.252.215   <none>        80:30816/TCP   9d

Static IP address

Create a static global IP address named smarp-ip

gcloud compute addresses create smarp-ip --global
gcloud compute addresses describe smarp-ip --global

DNS record

In order to have a domain / sub-domain (e.g. smarp.mirai-solutions.ch) pointing to the static smarp-ip address, domain name records must be configured with the domain registrar, by adding an A (Address) type DNS record for your domain or sub-domain name and have its value configured with the reserved IP address.

Once this is done you should be able to visit the app by typing the domain / sub-domain as URL (e.g. https://smarp.mirai-solutions.ch). It might take several hours for the DNS records to propagate, you can check it via

host smarp.mirai-solutions.ch
## smarp.mirai-solutions.ch has address 35.186.252.175

TLS certificate for HTTPS

Google Cloud Platform offers Google-managed TLS certificates for (sub)-domains enabling HTTPS traffic. Certificates are automatically provisioned and renewed and can be specified in an Ingress.

Note that the managed certificates are in a Beta release at the time of writing.

See How-to: Using Google-managed SSL certificates.

Kubernetes cluster update

Note: Managed certificates require clusters with masters running Kubernetes 1.12.6-gke.7 or higher.

You can check the serverVersion via

kubectl version -o=yaml | grep -i version

If the version is below 1.12.6-gke.7, You will have to manually upgrade the cluster

To find the supported Kubernetes master and node versions for upgrades and downgrades, run the following command:

gcloud container get-server-config

You can manually upgrade the cluster’s master, and then the nodes can be upgraded to the same version.

To upgrade the master’s version of the smarp cluster to a specific version (that is not the default) run e.g.

gcloud container clusters upgrade smarp --master --cluster-version 1.12.6-gke.10

where 1.12.6-gke.10 is a new-enough version (typically the latest) you found listed above. This can take 10-20 minutes.

Once you’ve upgraded your cluster’s master, the nodes can be upgraded to the same version. The following command upgrades your nodes to the version that your master is running:

gcloud container clusters upgrade smarp

Wait for the operation to complete (a few minutes).

Google-managed certificate for the Ingress

You can define a ManagedCertificate resource for the smarp.mirai-solutions.ch domain from above (pointing to smarp-ip) using the YAML manifest

## # smarp-certificate.yaml
## apiVersion: networking.gke.io/v1beta1
## kind: ManagedCertificate
## metadata:
##   name: smarp-certificate
## spec:
##   domains:
##   - smarp.mirai-solutions.ch

and create the corresponding ManagedCertificate resource

kubectl apply -f smarp-certificate.yaml

It is going to take between 10 minutes and up to 2 hours before the certificate is provisioned and activated. While this occurs, you can still proceed to creating the Ingress below, possibly watching the Certificate Status via

watch kubectl describe managedcertificate smarp-certificate

It will eventually switch from status Provisioning to status Active, while the Events section will switch from Create to <none>. Note that the Domain Status can switch to Status: FailedNotVisible along the way, especially if the TTL of your DNS record is low. This doesn’t prevent the certificate from activating, however if the event switches away from Create and you still see Certificate Status: Provisioning, you may have to delete the certificate and check if anything is wrong with your DNS configuration.

Ingress configuration

Create the Ingress exposing the app via an external IP address, associated with the TLS certificate defined above.

In the manifest, you can see that incoming requests are routed to port 80 of the NodePort service backend smarp-backend. Specify smarp-ip in the Ingress manifest as global-static-ip-name annotation, and also add smarp-certificate as networking.gke.io/managed-certificates

## # smarp-ingress.yaml
## apiVersion: extensions/v1beta1
## kind: Ingress
## metadata:
##   name: smarp-ingress
##   annotations:
##     kubernetes.io/ingress.global-static-ip-name: smarp-ip
##     networking.gke.io/managed-certificates: smarp-certificate
##     kubernetes.io/ingress.allow-http: "false"
## spec:
##   backend:
##     serviceName: smarp-backend
##     servicePort: 80

Then create the corresponding Ingress resource

kubectl apply -f smarp-ingress.yaml

Wait a few minutes for the Ingress controller to configure an HTTP(S) load balancer and an associated backend service. Be patient (8-12 minutes):

Note: It may take a few minutes for GKE to allocate an external IP address and set up forwarding rules until the load balancer is ready to serve your application. In the meanwhile, you may get errors such as HTTP 404 or HTTP 500 until the load balancer configuration is propagated across the globe.

You can watch the status until you eventually see an IP address being allocated to the Ingress

watch kubectl get ingress smarp-ingress

You should be able to reach the app via HTTPS https://smarp.mirai-solutions.ch, note that this can take longer and require another 10-15 minutes until the certificate is active (see above). You may want to keep watching it via

watch -n 30 curl https://smarp.mirai-solutions.ch/

Disable HTTP traffic

Note that we have disabled the HTTP traffic by including the annotation kubernetes.io/ingress.allow-http: "false" in the YAML manifest.

Note that annotations can also be added or removed dynamically, e.g.

kubectl annotate ingress smarp-ingress kubernetes.io/ingress.allow-http-

Notice the minus sign, -, at the end of the command.

References

OBSOLETE



miraisolutions/SmaRP documentation built on March 23, 2022, 6:46 a.m.