RH Summit 2020 Demo Cluster Setup

This repo contains all scripts to prepare an OpenShift 4 cluster to run the 2020 Live Keynote Demo.

Setup

Required tooling includes oc to deploy to OpenShift, yq for Kafka, and skupper for the messaging interconnect.

Create a .env file with all the required environment variables. An example env file .env.example is included.

cp .env.example .env

Test that you can login to the server with

make oc_login

Strimzi and Apache Kafka

The Apache Kafka related deployment is made by:

• The Strimzi operators, starting from the Cluster operator to the Topic and User operators
• The Apache Kafka cluster deployment (alongside with a Zookeeper cluster)
• The monitoring infrastructure made by Prometheus, the related alert manager and Grafana with Kafka and Zookeeper dashboards

To deploy the Apache Kafka infrastructure:

make kafka

In order to deploy Kafka, this version of the yq tool is needed.

HQ Applications

Kafka

This will create the kafka-demo project. See above for additional info. Before deploying the Kafka cluster on HQ, change the number of partitions in the KafkaTopic (in the kafka/cluster/kafka-topics.yaml folder) to 300 because it will contain an aggregation of messages coming from the clusters at the edge.

make kafka

When the Kafka cluster is deployed, the next step is to deploy Kafka Mirror Maker 2 for mirroring data from all the clusters at the edge. Before running the deployment, be sure that the clusters.lbs file exists and it contains the Load Balancer addresses for all the Kafka clusters already deployed at the edge that have to be mirrored to the HQ.

make kafka_mirror_maker

Datagrid Service

This will create the datagrid-demo project and 2 pods for the datagrid service. Provides storage for the game state and bot config which will be pushed to the edge clusters.

make datagrid

Admin UI and Services

• admin-server 2 pods. Service which sets global settings and sends AMQP messages to the edge applications.
• admin-ui 2 pods. Service for the Admin UI application which connects to the admin-server internally. Exposed via route.
• dashboard-ui 2 pods. Service for the UI application which shows the 3d rendered global map of clusters.
• skupper service interconnects across all edges and admin.

To configure an admin username and password, set the ADMIN_USERNAME and ADMIN_PASSWORD in the .env file. for the frontend to show the currently connected cluster, set the CLUSTER_NAME in the .env file.

make admin

Admin UI: http://ui-admin-hq.hub-cluster (blank username/password unless set in the .env file)

Leaderboard Services

Creates services to consume data from mirrored kafka streams and aggregate and store them in Postgres. Also deploys services to allow access to that data from the Leaderboard UI.

make leaderboard_project
make leaderboard_install_postgresql
make leaderboard_install_all

Leaderboard UI

Creates the leaderboard project if not created and the UI application which polls the leaderboard services for display.

make visualization

Edge Applications

Edge clusters for the 2020 demonstration utilized clusters from IBM ROKS, Azure, GCP, and AWS in San Francisco, New York, Sao Paulo, London, Frankfurt, Singapore, and Sydney. Skupper should be available on Admin-HQ before installing edge.

The Edge clusters require

• datagrid
• kafka
• digit recognition service
• scoring service
• frontend applications
  • admin-edge
  • bot-server
  • phone-server
  • phone-ui
  • skupper

Kafka

This will create the kafka-demo project. See above for additional info. Before deploying the Kafka cluster on HQ, be sure that the number of partitions in the KafkaTopic (in the kafka/cluster/kafka-topics.yaml folder) is the 50 default value.

make kafka

When the Kafka cluster is deployed, the script gets the Load Balancer address and append it into a clusters.lbs file. This file will contains all the Load Balancer addresses related to the edge clusters where Kafka is deployed that have to be mirrored to the HQ.

Datagrid Service

This will create the datagrid-demo project and 2 pods for the datagrid service. Provides edge storage as a cache with notifications to frontend services.

make datagrid

London is backed-up to Frankfurt via remote store and additional scripting is done for those clusters. For detailed deployment instructions please see the Data Grid README.

Digit Recognition Service

This will create the ai project and 12 pods for the digit-recognition service.

make ml

Scoring Service

This will create the scoring project and 5 pods for the scoring service.

make quarkus-scoring

Front End UI and Services

This will create the frontend project and services. For the frontend to show the currently connected cluster, set the CLUSTER_NAME in the .env file.

• admin-edge 2 pods. Service which listens for AMQP messages from the global admin application and writes to Infinispan
• bot-server 2 pods. Service which plays the game for testing purposes
• phone-server 5 pods. Service for the game accessible via websocket.
• phone-ui 5 pods. Service for the UI application which connects to the phone-server internally. Also exposed via route.
• skupper service interconnects across all edges and admin.

make frontend

Game is accessible via http://game-frontend.edge-host

Adding deployment scripts

Just create a sub directory, add a shell script and resource files and execute them via the Makefile. Note that the commands in a Makefile have to be indented by tabs. Also add a short description to this README.md