deployment

Kubernetes Deployment

This document will overview the dapr-pipeline demo deployment into Kubernetes. For illustration purposes, all commands in this document will be based on Microsoft Azure.

Dapr supports a wide array of state and pubsub backing services across multiple Cloud and on-prem deployments, so if you already have a Kubernates cluster somewhere else, you can substitute:

• state backing service options
• pubsub backing service options

Prerequisite

• Azure CLI

Also, to simplify all the scripts in this doc, set a few az CLI defaults:

az account set --subscription <id or name>
az configure --defaults location=<prefered location> group=<your resource group>

Cluster (optional)

If you don't already have Kubernates cluster, you can create it on Azure with all the necessary add-ons for this demo usign tihs command:

az aks create --name demo \
              --kubernetes-version 1.15.10 \
              --enable-managed-identity \
              --vm-set-type VirtualMachineScaleSets \
              --node-vm-size Standard_F4s_v2 \
              --enable-addons monitoring,http_application_routing \
              --generate-ssh-keys

Install dapr

See instructions how to install dapr into your Kubernetes cluster here

Component-backing services

Assuming you have a Kubernates cluster and kubectl CLI configure to connect you are ready to setup the dapr components and their backing services:

This demo installs into the default namespace in your cluster. When installing into a different namespace, make sure to append the -n <your namespace name> to all commands below (secret, component, and deployment)

State

To configure dapr state component in this demo I will use Azure Table Storage. To set it up, follow these instructions. Once finished, you will need to configure also the Kubernates secrets to hold the Azure Table Storage account information:

kubectl create secret generic demo-state-secret \
  --from-literal=account-name='' \
  --from-literal=account-key=''

Once the secret is configured, deploy the dapr state component:

kubectl apply -f component/state.yaml

PubSub

To configure dapr pubsub component in this demo I will use Azure Service Bus. To set it up, follow these instructions. Once finished, you will need to configure the Kubernates secret to hold Azure Service Bus connection string information.

kubectl create secret generic demo-pubsub-secret \
  --from-literal=connection-string=''

Once the secret is configured, deploy the dapr pubsub topic components:

kubectl apply -f component/pubsub.yaml

Twitter Input Binding

Finally, to use the Dapr Twitter input binding we need to configure the Twitter API secretes. You can get these by registering Twitter application and obtain this information here.

kubectl create secret generic demo-twitter-secrets \
  --from-literal=access-secret="" \
  --from-literal=access-token="" \
  --from-literal=consumer-key="" \
  --from-literal=consumer-secret=""

Once the secret is configured you can deploy the Twitter binding:

kubectl apply -f component/twitter.yaml

Deploying Demo

Once the necessary components are created, you just need to create one more secret for the Cognitive Service token that is used in the processor service:

kubectl create secret generic demo-sentimenter-secret \
  --from-literal=token=""

And now you can deploy the entire pipeline (provider, processor, viewer) with a single command:

kubectl apply -f sentimenter.yaml -f viewer.yaml -f processor.yaml

You can check on the status of your deployment like this:

kubectl get pods -l env=demo

The results should look similar to this (make sure each pod has READY status 2/2)

NAME                           READY   STATUS    RESTARTS   AGE
processor-89666d54b-hkd5t      2/2     Running   0          18s
sentimenter-85cfbf5456-lc85g   2/2     Running   0          18s
viewer-76448d65fb-bm2dc        2/2     Running   0          18s

Exposing viewer UI

To expose the viewer application externally, create Kubernetes service using deployment/service/viewer.yaml

kubectl apply -f service/viewer.yaml

Note, the provisioning of External IP may take up to 1-2 min

To view the viewer application by capturing the load balancer public IP and opening it in the browser:

export VIEWER_IP=$(kubectl get svc viewer --output 'jsonpath={.status.loadBalancer.ingress[0].ip}')
open "http://${VIEWER_IP}/"

To change the Twitter topic query, first edit the deployment/component/twitter.yaml, then apply it (kubectl apply -f component/twitter.yaml), and finally, restart the processor (kubectl rollout restart deployment processor) to ensure the new configuration is applied.

Observability

Azure

You can view the scored tweets in Azure table storage

Similarly you can monitor the pubsub topic throughout in Azure Service Bus

OSS

In addition, you can also observe Dapr metrics, logs, and traces for this demo.

Metrics in Grafana dashboard

Instructions on how to setup Grafana for Dapr are here

Forward port

http://localhost:8080/

Logs in Kibana dashboard

Instructions on how to setup Kibana for Dapr are here

http://localhost:5601/

Traces in Zipkin dashboard

Instructions on how to setup Zipkin for Dapr are here

http://localhost:9411/zipkin/

Note, if your Zipkin isn't deployed in the default namespace you will have to edit the exporterAddress in deployment/tracing/zipkin.yaml

Then just restart all the deployments

kubectl rollout restart deployment processor sentimenter  viewer